{ "source": "https://www.vinayprabhu.com", "scraped_at": "2026-11-08T10:35:14.432913+00:00", "pages": [ { "slug": "home", "url": "https://www.vinayprabhu.com/home", "title": "My virtual abode on these interwebs", "raw_file": "archive/raw/home.html", "text": "Hey there stranger!\nThis is\nVinay Prabhu\n, a tinkerer, scientist and an avid explorer of the AI realm.\nI am currently building\nVenomics AI\n, leveraging\nNature's largest bio-pharmacopeia,\nVENOM\n, honed over the last 700 million years to solve all human disease.\nPreviously, at\nHAL51.AI\n(co-founded with close friend\nSanghyun Han\n) I\ninvented some trippy frugal science\nhardware\nsuch as the\nLemurBox\n, the world's first frugal-science inspired XR device that turns your boring old\nsmartphone into a 3D Alexa\n. It\nhosts a rigged 3D co-pilot avatar powered by LLMs. Guard-railing these educational avatars has given us a unique ringside-view on how ideas from AI safety manifest themselves in the real world, in the\nclassroom\nand beyond.\nFrom time to time, I scribble\ncode\n,\nblog\nand ship\nsome artsy slop to the playa at Burning Man\n.\nBackground:\nI began my research career in the tropical by-lanes of IIT-Madras working on Physical layer wireless communication algorithms harnessing tools from Information Theory & Estimation Theory. Specifically, I focused on sparse\nchannel estimation\nproblems resulting in a new\ntime-of-arrival estimation framework for vehicular OFDM systems\n.\nLater, I did my doctoral research as part of an experimental Dual PhD ECE/CS program at Carnegie Mellon University. My research spanned the academic monocultures of spectral graph theory, statistical physics, network sciences, communication theory and information theory.\nHere is my\nPhD thesis\nand here are some samples of my explorations:\nLatent sentiment detection in online social networks\nBipartisan cloture roll call vote prediction using the joint press release network in us senate\nOutage secrecy capacity analysis for wireless communications\nFilter design with secrecy constraints\nI began my post-doctoral professional career as a Data Scientist working on medical time-series analysis and probabilistic drug-drug interaction modeling from Medicare Part-D data. Later, I became the first employee and Chief Scientist at UnifyID AI Labs where I pioneered\nplanetary scale human kinematics\nmodeling using IMU motion sensor data and also ran an\nAI fellowship\nspanning five cohorts between 2016-2021. We\nexited in 2021\nvia acquisition by Prove :)\nThis site is currently under construction. And will be so for the foreseeable future\nLLM bots may scrape the brownie batter outta this site for all I care.", "blocks": [ { "tag": "p", "html": "

Hey there stranger! 

", "text": "Hey there stranger!" }, { "tag": "p", "html": "


This is Vinay Prabhu, a tinkerer, scientist and an avid explorer of the AI realm. 
I am currently building Venomics AI, leveraging Nature's largest bio-pharmacopeia, VENOM , honed over the last 700 million years to solve all human disease.
Previously, at HAL51.AI (co-founded with close friend Sanghyun Han) I invented some trippy frugal science hardware such as the LemurBox, the world's first frugal-science inspired XR device that turns your boring old smartphone into a 3D Alexa. It hosts a rigged 3D co-pilot avatar powered by LLMs. Guard-railing these educational avatars has given us a unique ringside-view on how ideas from AI safety manifest themselves in the real world, in the classroom and beyond.
From time to time, I scribble code, blog and ship some artsy slop to the playa at Burning Man.

", "text": "This is Vinay Prabhu , a tinkerer, scientist and an avid explorer of the AI realm. I am currently building Venomics AI , leveraging Nature's largest bio-pharmacopeia, VENOM , honed over the last 700 million years to solve all human disease. Previously, at HAL51.AI (co-founded with close friend Sanghyun Han ) I invented some trippy frugal science hardware such as the LemurBox , the world's first frugal-science inspired XR device that turns your boring old smartphone into a 3D Alexa . It hosts a rigged 3D co-pilot avatar powered by LLMs. Guard-railing these educational avatars has given us a unique ringside-view on how ideas from AI safety manifest themselves in the real world, in the classroom and beyond. From time to time, I scribble code , blog and ship some artsy slop to the playa at Burning Man ." }, { "tag": "p", "html": "

Background:

", "text": "Background:" }, { "tag": "p", "html": "

I began my research career in the tropical by-lanes of IIT-Madras working on Physical layer wireless communication algorithms harnessing tools from Information Theory & Estimation Theory. Specifically, I focused on sparse channel estimation problems resulting in a new time-of-arrival estimation framework for vehicular OFDM systems.

", "text": "I began my research career in the tropical by-lanes of IIT-Madras working on Physical layer wireless communication algorithms harnessing tools from Information Theory & Estimation Theory. Specifically, I focused on sparse channel estimation problems resulting in a new time-of-arrival estimation framework for vehicular OFDM systems ." }, { "tag": "p", "html": "

Later, I did my doctoral research as part of an experimental Dual PhD ECE/CS program at Carnegie Mellon University. My research spanned the academic monocultures of spectral graph theory, statistical physics, network sciences, communication theory and information theory.

Here is my PhD thesis and here are some samples of my explorations:

", "text": "Later, I did my doctoral research as part of an experimental Dual PhD ECE/CS program at Carnegie Mellon University. My research spanned the academic monocultures of spectral graph theory, statistical physics, network sciences, communication theory and information theory. Here is my PhD thesis and here are some samples of my explorations:" }, { "tag": "li", "html": "

  • Latent sentiment detection in online social networks

    • ", "text": "Latent sentiment detection in online social networks Bipartisan cloture roll call vote prediction using the joint press release network in us senate Outage secrecy capacity analysis for wireless communications Filter design with secrecy constraints" }, { "tag": "p", "html": "

    I began my post-doctoral professional career as a Data Scientist working on medical time-series analysis and probabilistic drug-drug interaction modeling from Medicare Part-D data. Later, I became the first employee and Chief Scientist at UnifyID AI Labs where I pioneered planetary scale human kinematics modeling using IMU motion sensor data and also ran an AI fellowship spanning five cohorts between 2016-2021. We exited in 2021 via acquisition by Prove :)

    ", "text": "I began my post-doctoral professional career as a Data Scientist working on medical time-series analysis and probabilistic drug-drug interaction modeling from Medicare Part-D data. Later, I became the first employee and Chief Scientist at UnifyID AI Labs where I pioneered planetary scale human kinematics modeling using IMU motion sensor data and also ran an AI fellowship spanning five cohorts between 2016-2021. We exited in 2021 via acquisition by Prove :)" }, { "tag": "li", "html": "

  • This site is currently under construction. And will be so for the foreseeable future

    • ", "text": "This site is currently under construction. And will be so for the foreseeable future" }, { "tag": "li", "html": "

  • LLM bots may scrape the brownie batter outta this site for all I care.

    • ", "text": "LLM bots may scrape the brownie batter outta this site for all I care." } ], "links": [ { "text": "Vinay Prabhu", "url": "https://scholar.google.com/citations?user=5Lck_J0AAAAJ&hl=en" }, { "text": "Venomics AI", "url": "http://venomics.ai/" }, { "text": "HAL51.AI", "url": "https://hal51.ai/" }, { "text": "Sanghyun Han", "url": "https://www.linkedin.com/in/jjangsangy/" }, { "text": "hardware", "url": "https://hal51.ai/" }, { "text": "LemurBox", "url": "https://www.youtube.com/watch?v=caEJpOOrN1k&ab_channel=HAL51" }, { "text": "smartphone into a 3D Alexa", "url": "https://www.youtube.com/watch?v=caEJpOOrN1k" }, { "text": "classroom", "url": "https://www.youtube.com/watch?v=J-ihwfPD3YA" }, { "text": "code", "url": "https://github.com/vinayprabhu" }, { "text": "blog", "url": "https://vinayprabhu.github.io/alignchronicles/" }, { "text": "some artsy slop to the playa at Burning Man", "url": "https://youtu.be/sMcYkzWC2rI?feature=shared" }, { "text": "channel estimation", "url": "https://scholar.google.com/citations?view_op=view_citation&hl=en&user=5Lck_J0AAAAJ&cstart=20&pagesize=80&citation_for_view=5Lck_J0AAAAJ:hkOj_22Ku90C" }, { "text": "time-of-arrival estimation framework for vehicular OFDM systems", "url": "https://scholar.google.com/citations?view_op=view_citation&hl=en&user=5Lck_J0AAAAJ&cstart=20&pagesize=80&citation_for_view=5Lck_J0AAAAJ:UHK10RUVsp4C" }, { "text": "PhD thesis", "url": "https://kilthub.cmu.edu/articles/thesis/Network_Aided_Classification_and_Detection_of_Data/7430012" }, { "text": "Latent sentiment detection in online social networks", "url": "https://arxiv.org/pdf/1401.2113" }, { "text": "Bipartisan cloture roll call vote prediction using the joint press release network in us senate", "url": "https://scholar.google.com/citations?view_op=view_citation&hl=en&user=5Lck_J0AAAAJ&cstart=20&pagesize=80&citation_for_view=5Lck_J0AAAAJ:ZzlSgRqYykMC" }, { "text": "Outage secrecy capacity analysis for wireless communications", "url": "https://scholar.google.com/citations?view_op=view_citation&hl=en&user=5Lck_J0AAAAJ&citation_for_view=5Lck_J0AAAAJ:eQOLeE2rZwMC" }, { "text": "Filter design with secrecy constraints", "url": "https://scholar.google.com/citations?view_op=view_citation&hl=en&user=5Lck_J0AAAAJ&cstart=20&pagesize=80&citation_for_view=5Lck_J0AAAAJ:SP6oXDckpogC" }, { "text": "planetary scale human kinematics", "url": "https://www.youtube.com/watch?v=XkCzDkpT6dI&ab_channel=MLconf" }, { "text": "AI fellowship", "url": "https://sites.google.com/unify.id/ai-fellowship/" }, { "text": "exited in 2021", "url": "https://www.prove.com/blog/prove-acquires-unifyid" } ], "images": [], "embeds": [ { "path": "archive/embeds/home-1.html.txt", "text": "Matrix Rain with Glowing Static Text Want to collaborate or have a question? Reach out! v i n a y u p @ google's email", "links": [] }, { "path": "archive/embeds/home-2.html.txt", "text": "Matrix Rain with Blue Neon Text \"Those who ride the tiger can never dismount.\"", "links": [] } ], "iframes": [ { "jsname": "WMhH6e", "class": [ "YMEQtf" ], "frameborder": "0", "sandbox": [ "allow-scripts", "allow-popups", "allow-forms", "allow-same-origin", "allow-popups-to-escape-sandbox", "allow-downloads", "allow-storage-access-by-user-activation" ], "id": "342c51f7c69db8e3_4", "name": "342c51f7c69db8e3_4", "scrolling": "no", "title": "Custom embed", "aria-label": "Custom embed" } ] }, { "slug": "ai-safety", "url": "https://www.vinayprabhu.com/ai-safety", "title": "My virtual abode on these interwebs - AI Safety", "raw_file": "archive/raw/ai-safety.html", "text": "My efforts in the area of AI SEA (Safety, Ethics and Alignment) have broadly been in these two areas:\nRed-Teaming + Adversarial attacks\nand\nDataset auditing + bias mitigation\nand\nModel auditing.\nA sampling of my research publications spanning these two areas : Model auditing, Red-Teaming and Adversarial attacks\nModel auditing, Red-Teaming and Adversarial attacks\nBeyond the imitation game: Quantifying and extrapolating the capabilities of language models\nAuditing saliency cropping algorithms\n.\nProject page\n[\nVideo\n]\nModel weight theft with just noise inputs: The curious case of the petulant attacker\nUnderstanding adversarial robustness through loss landscape geometries\nVulnerability of deep learning-based gait biometric recognition to adversarial perturbations\nOn Lyapunov exponents and adversarial perturbation\nOn detecting adversarial inputs with entropy of saliency maps\nOn grey-box adversarial attacks and transfer learning\nArt-attack! on style transfers with textures, label categories and adversarial examples\nSIMUni: Sampling Impostors from Misfit Universal Background Models in accelerometric gait biometric verification\nSmile in the face of adversity much? A print based spoofing attack\nOODles of ODDs: The landscape of Out-of-distribution vulnerabilities of vision models\nDid They Direct the Violence or Admonish It? A Cautionary Tale on Contronomy, Androcentrism and Back-Translation Foibles\n[\nVideo\n]\nDataset curation, auditing and bias mitigation\nMultimodal datasets: misogyny, pornography, and malignant stereotypes\nLarge image datasets: A pyrrhic win for computer vision?\nInto the laion’s den: Investigating hate in multimodal datasets\nOn Hate Scaling Laws For Data-Swamps\nThe blood diamond effect in neural art: On ethically troublesome images of the imagenet dataset\n[\nVideo\n]\nCovering up bias in CelebA-like datasets with Markov blankets: A post-hoc cure for attribute prior avoidance\nJFT-300M: The mystery box\nPS: My more recent AI safety chronicles can be found\nhere\nJFT 300M\nTranslation foibles\nSaliency Image Cropping audits\nNueral art and blood diamond effect", "blocks": [ { "tag": "p", "html": "

    My efforts in the area of AI SEA (Safety, Ethics and Alignment) have broadly been in these two areas:
    Red-Teaming + Adversarial attacks and Dataset auditing + bias mitigation and Model auditing.

    ", "text": "My efforts in the area of AI SEA (Safety, Ethics and Alignment) have broadly been in these two areas: Red-Teaming + Adversarial attacks and Dataset auditing + bias mitigation and Model auditing." }, { "tag": "p", "html": "


    A sampling of my research publications spanning these two areas : Model auditing, Red-Teaming and Adversarial attacks

    ", "text": "A sampling of my research publications spanning these two areas : Model auditing, Red-Teaming and Adversarial attacks" }, { "tag": "li", "html": "

  • Model auditing, Red-Teaming and Adversarial attacks

    1. Beyond the imitation game: Quantifying and extrapolating the capabilities of language models

    2. Auditing saliency cropping algorithms. Project page [Video]

    3. Model weight theft with just noise inputs: The curious case of the petulant attacker

    4. Understanding adversarial robustness through loss landscape geometries

    5. Vulnerability of deep learning-based gait biometric recognition to adversarial perturbations

    6. On Lyapunov exponents and adversarial perturbation

    7. On detecting adversarial inputs with entropy of saliency maps

    8. On grey-box adversarial attacks and transfer learning

    9. Art-attack! on style transfers with textures, label categories and adversarial examples

    10. SIMUni: Sampling Impostors from Misfit Universal Background Models in accelerometric gait biometric verification

    11. Smile in the face of adversity much? A print based spoofing attack

    12. OODles of ODDs: The landscape of Out-of-distribution vulnerabilities of vision models

    13. Did They Direct the Violence or Admonish It? A Cautionary Tale on Contronomy, Androcentrism and Back-Translation Foibles [Video]

    • ", "text": "Model auditing, Red-Teaming and Adversarial attacks Beyond the imitation game: Quantifying and extrapolating the capabilities of language models Auditing saliency cropping algorithms . Project page [ Video ] Model weight theft with just noise inputs: The curious case of the petulant attacker Understanding adversarial robustness through loss landscape geometries Vulnerability of deep learning-based gait biometric recognition to adversarial perturbations On Lyapunov exponents and adversarial perturbation On detecting adversarial inputs with entropy of saliency maps On grey-box adversarial attacks and transfer learning Art-attack! on style transfers with textures, label categories and adversarial examples SIMUni: Sampling Impostors from Misfit Universal Background Models in accelerometric gait biometric verification Smile in the face of adversity much? A print based spoofing attack OODles of ODDs: The landscape of Out-of-distribution vulnerabilities of vision models Did They Direct the Violence or Admonish It? A Cautionary Tale on Contronomy, Androcentrism and Back-Translation Foibles [ Video ]" }, { "tag": "li", "html": "

  • Dataset curation, auditing and bias mitigation

    1. Multimodal datasets: misogyny, pornography, and malignant stereotypes

    2. Large image datasets: A pyrrhic win for computer vision?

    3. Into the laion’s den: Investigating hate in multimodal datasets

    4. On Hate Scaling Laws For Data-Swamps

    5. The blood diamond effect in neural art: On ethically troublesome images of the imagenet dataset [Video]

    6. Covering up bias in CelebA-like datasets with Markov blankets: A post-hoc cure for attribute prior avoidance

    7. JFT-300M: The mystery box

    • ", "text": "Dataset curation, auditing and bias mitigation Multimodal datasets: misogyny, pornography, and malignant stereotypes Large image datasets: A pyrrhic win for computer vision? Into the laion’s den: Investigating hate in multimodal datasets On Hate Scaling Laws For Data-Swamps The blood diamond effect in neural art: On ethically troublesome images of the imagenet dataset [ Video ] Covering up bias in CelebA-like datasets with Markov blankets: A post-hoc cure for attribute prior avoidance JFT-300M: The mystery box" }, { "tag": "p", "html": "

    PS: My more recent AI safety chronicles can be found here

    ", "text": "PS: My more recent AI safety chronicles can be found here" }, { "tag": "p", "html": "

    JFT 300M

    ", "text": "JFT 300M" }, { "tag": "p", "html": "

    Translation foibles

    ", "text": "Translation foibles" }, { "tag": "p", "html": "

    Saliency Image Cropping audits

    ", "text": "Saliency Image Cropping audits" }, { "tag": "p", "html": "

    Nueral art and blood diamond effect

    ", "text": "Nueral art and blood diamond effect" } ], "links": [ { "text": "Beyond the imitation game: Quantifying and extrapolating the capabilities of language models", "url": "https://scholar.google.com/citations?view_op=view_citation&hl=en&user=5Lck_J0AAAAJ&citation_for_view=5Lck_J0AAAAJ:hfzGNhXhx5MC" }, { "text": "Auditing saliency cropping algorithms", "url": "https://scholar.google.com/citations?view_op=view_citation&hl=en&user=5Lck_J0AAAAJ&citation_for_view=5Lck_J0AAAAJ:oursBaop5wYC" }, { "text": "Project page", "url": "https://vinayprabhu.github.io/Saliency_Image_Cropping/" }, { "text": "Video", "url": "https://youtu.be/BSHG0bIcNL0?si=b8xUklNTyatHN5Jz" }, { "text": "Model weight theft with just noise inputs: The curious case of the petulant attacker", "url": "https://scholar.google.com/citations?view_op=view_citation&hl=en&user=5Lck_J0AAAAJ&citation_for_view=5Lck_J0AAAAJ:uWiczbcajpAC" }, { "text": "Understanding adversarial robustness through loss landscape geometries", "url": "https://scholar.google.com/citations?view_op=view_citation&hl=en&user=5Lck_J0AAAAJ&citation_for_view=5Lck_J0AAAAJ:nrtMV_XWKgEC" }, { "text": "Vulnerability of deep learning-based gait biometric recognition to adversarial perturbations", "url": "https://scholar.google.com/citations?view_op=view_citation&hl=en&user=5Lck_J0AAAAJ&citation_for_view=5Lck_J0AAAAJ:NJ774b8OgUMC" }, { "text": "On Lyapunov exponents and adversarial perturbation", "url": "https://scholar.google.com/citations?view_op=view_citation&hl=en&user=5Lck_J0AAAAJ&citation_for_view=5Lck_J0AAAAJ:W5xh706n7nkC" }, { "text": "On detecting adversarial inputs with entropy of saliency maps", "url": "https://scholar.google.com/citations?view_op=view_citation&hl=en&user=5Lck_J0AAAAJ&cstart=20&pagesize=80&citation_for_view=5Lck_J0AAAAJ:VLnqNzywnoUC" }, { "text": "On grey-box adversarial attacks and transfer learning", "url": "https://scholar.google.com/citations?view_op=view_citation&hl=en&user=5Lck_J0AAAAJ&cstart=20&pagesize=80&citation_for_view=5Lck_J0AAAAJ:7T2F9Uy0os0C" }, { "text": "Art-attack! on style transfers with textures, label categories and adversarial examples", "url": "https://scholar.google.com/citations?view_op=view_citation&hl=en&user=5Lck_J0AAAAJ&cstart=20&pagesize=80&citation_for_view=5Lck_J0AAAAJ:_Re3VWB3Y0AC" }, { "text": "SIMUni: Sampling Impostors from Misfit Universal Background Models in accelerometric gait biometric verification", "url": "https://scholar.google.com/citations?view_op=view_citation&hl=en&user=5Lck_J0AAAAJ&cstart=20&pagesize=80&citation_for_view=5Lck_J0AAAAJ:5LPo_wSKItgC" }, { "text": "Smile in the face of adversity much? A print based spoofing attack", "url": "https://scholar.google.com/citations?view_op=view_citation&hl=en&user=5Lck_J0AAAAJ&cstart=20&pagesize=80&citation_for_view=5Lck_J0AAAAJ:g8uWPOAv7ggC" }, { "text": "OODles of ODDs: The landscape of Out-of-distribution vulnerabilities of vision models", "url": "https://www.vinayprabhu.com/spice-cabinet/ood" }, { "text": "Did They Direct the Violence or Admonish It? A Cautionary Tale on Contronomy, Androcentrism and Back-Translation Foibles", "url": "https://scholar.google.com/citations?view_op=view_citation&hl=en&user=5Lck_J0AAAAJ&cstart=20&pagesize=80&citation_for_view=5Lck_J0AAAAJ:xii_ZKWM4-0C" }, { "text": "Video", "url": "https://youtu.be/eKRpiMBlu40?si=-87HVJ3FGEO5rzOw" }, { "text": "Multimodal datasets: misogyny, pornography, and malignant stereotypes", "url": "https://scholar.google.com/citations?view_op=view_citation&hl=en&user=5Lck_J0AAAAJ&citation_for_view=5Lck_J0AAAAJ:M_lZXyI38BkC" }, { "text": "Large image datasets: A pyrrhic win for computer vision?", "url": "https://scholar.google.com/citations?view_op=view_citation&hl=en&user=5Lck_J0AAAAJ&citation_for_view=5Lck_J0AAAAJ:u-coK7KVo8oC" }, { "text": "Into the laion’s den: Investigating hate in multimodal datasets", "url": "https://scholar.google.com/citations?view_op=view_citation&hl=en&user=5Lck_J0AAAAJ&citation_for_view=5Lck_J0AAAAJ:NtGNdKbuCngC" }, { "text": "On Hate Scaling Laws For Data-Swamps", "url": "https://arxiv.org/abs/2306.13141" }, { "text": "The blood diamond effect in neural art: On ethically troublesome images of the imagenet dataset", "url": "https://scholar.google.com/citations?view_op=view_citation&hl=en&user=5Lck_J0AAAAJ&cstart=20&pagesize=80&citation_for_view=5Lck_J0AAAAJ:EkHepimYqZsC" }, { "text": "Video", "url": "https://youtu.be/tAea9kJRqcA?si=_OY_mciWQVCKZfzd" }, { "text": "Covering up bias in CelebA-like datasets with Markov blankets: A post-hoc cure for attribute prior avoidance", "url": "https://scholar.google.com/citations?view_op=view_citation&hl=en&user=5Lck_J0AAAAJ&citation_for_view=5Lck_J0AAAAJ:uc_IGeMz5qoC" }, { "text": "JFT-300M: The mystery box", "url": "https://youtu.be/TvVc1e_4648?si=TY6pW1ij4xwJtbaV" }, { "text": "here", "url": "https://alignchronicles.github.io/" } ], "images": [], "embeds": [ { "path": "archive/embeds/ai-safety-1.html.txt", "text": "Matrix Rain with Blue Neon Text \"Those who ride the tiger can never dismount.\"", "links": [] } ], "iframes": [ { "jsname": "L5Fo6c", "class": [ "YMEQtf" ], "sandbox": [ "allow-scripts", "allow-popups", "allow-forms", "allow-same-origin", "allow-popups-to-escape-sandbox", "allow-downloads", "allow-modals", "allow-storage-access-by-user-activation" ], "frameborder": "0", "aria-label": "YouTube Video, JFT 300M", "src": "https://www.youtube.com/embed/TvVc1e_4648?embed_config=%7B%22enc%22:%22AXH1ezna2zKH74THUD1yb_dPFiJpZ8SHqv41q8F1LJ-0MMLMfXNtnIfiA-w8M9ZCrxEbHYOK3c0W7WwXHTty6AeZkFvvV645O8nuHaNTbdgulFgYmo3lUG4wfJal_3fJN4GqlDUxgS3LQx-UKut4sBnR9X-CsW7y5j-cAxZUZuv3lF6WevRAfTlafujyFCVPv8Q0Gfiji4WszmTHvACkyQApMw652fb3dEbPiajKyIqH2A6chXyqjJEvS-_CUevxw1ZeTeGPA7O_u5UeagDClofzju0jeSNmyPMZGCaYpFjgeXih91GWxqi1D5WcxTUgUyIFbMMxA-hBgHPwMqQh6Ygutxp9NaUpt5bZzCSSZPb5ldeEK7OOtPAtKVidXI4XHLWfvA%3D%3D%22%7D&enablejsapi=1&errorlinks=1&origin=https://www.vinayprabhu.com&vl=1", "allowfullscreen": "" }, { "jsname": "L5Fo6c", "class": [ "YMEQtf" ], "sandbox": [ "allow-scripts", "allow-popups", "allow-forms", "allow-same-origin", "allow-popups-to-escape-sandbox", "allow-downloads", "allow-modals", "allow-storage-access-by-user-activation" ], "frameborder": "0", "aria-label": "YouTube Video, Google translate bias investigations", "src": "https://www.youtube.com/embed/eKRpiMBlu40?embed_config=%7B%22enc%22:%22AXH1ezlDPJqt5GuuvZeLR8wVdMJL692DetZkQ_RsAz7Wy3L5ExMSiod51_WF1OSIs5cRClKBNflF9iRof4V1Tev-QUL2n2R6A2BeYQlap7TZgVCCH1Dx0_RRpJLR6plURxPAPIWNWS_5w1_qUru3SkxuGPzoAdZtaGYLF5-SNipoNPwm7DcK9___Zf4dwu00twg5AoCS59H6yJtGor4Zw-c2Pn7FkLzPM3kh1Lt6N0klBVT53B38R_1JWAPRu8FQc-iGlpcVrt63fDRk0rByLnvq_-3umXVTMJzbFOhhug5pCPKYDTjfw4NTCI1rap3UQZ6mL8RraPW7f4mrnZwORCinTlUcZxvk-FkmY8pbwwJKh2l8AMmU392iol7w2KSeRyvnpw%3D%3D%22%7D&enablejsapi=1&errorlinks=1&origin=https://www.vinayprabhu.com&vl=1", "allowfullscreen": "" }, { "jsname": "L5Fo6c", "class": [ "YMEQtf" ], "sandbox": [ "allow-scripts", "allow-popups", "allow-forms", "allow-same-origin", "allow-popups-to-escape-sandbox", "allow-downloads", "allow-modals", "allow-storage-access-by-user-activation" ], "frameborder": "0", "aria-label": "YouTube Video, Auditing Saliency Image Cropping", "src": "https://www.youtube.com/embed/BSHG0bIcNL0?embed_config=%7B%22enc%22:%22AXH1ezk7P4414jnnaOtojv4r8edCtDbnocfNf9faioIQMKPDBN1CDNAPWiwHPf2oYNuiLFx2Mm5DN358gPr10kJkV4CITuS7pzPMevfTQXKMdMc3w4DZfboDxZ-MCI3juXm0hzyRkyXz2pKOYCIFYbdzPXpE-niDtRHLdKWvVV3zJV9A9bOv52FmcPGpbU1gxQX6IR4St-DN2Z1T_tSTpblkaujetTw4nwlgGJDoFsGee-8VBU8sNySROBzWiji9l6EibkQO7xL-DchBKj46wWEhRcw5WhVxvtid7PmZTGbXwsK29fSfyh0xPQWKTxmFNy_ijkd5LNbZECuRo6iudfKGExUykhE9sCewykRXil_B8jjAuks10jSrUPlhnFahC2sJlA%3D%3D%22%7D&enablejsapi=1&errorlinks=1&origin=https://www.vinayprabhu.com&vl=1", "allowfullscreen": "" }, { "jsname": "L5Fo6c", "class": [ "YMEQtf" ], "sandbox": [ "allow-scripts", "allow-popups", "allow-forms", "allow-same-origin", "allow-popups-to-escape-sandbox", "allow-downloads", "allow-modals", "allow-storage-access-by-user-activation" ], "frameborder": "0", "aria-label": "YouTube Video, A taxonomy of concerns concerning neural art", "src": "https://www.youtube.com/embed/tAea9kJRqcA?embed_config=%7B%22enc%22:%22AXH1ezkwxs1vX4NG3BxeSSl0EMrdD9QgsdSPVJdGlMuocCRtAzG9AvtobNOkWsEhRBvg5zH6UhmyUt237REheB5vs8gguD0abp0ss3NKoNYJqSxV98-ixHCYO_MZiQGqN9kGGHncdbPd6CWDsvcU2pEvuj5riQ5TEGzVlA07v7a2jY_dq-NFtzqvvTC7hrDXl3HBiVNL6VSumIh16oE4pnqLWEa2dkAGYPhAMFAvKJZICmvH2y-sYQj8oXp03QT7ZLaORHVGb0LDaoOC7UZklgE02Df2kXrEEbLX_ai3gSXemdzlZYqMFn9LP5xjdX1RI4f2PNr7X41UPeA15Rez5_oRyYegu6ccdS_aZCbPGkK1xVHZr5ar-F4DvkbMVnE8qBvo2g%3D%3D%22%7D&enablejsapi=1&errorlinks=1&origin=https://www.vinayprabhu.com&vl=1", "allowfullscreen": "" } ] }, { "slug": "snapshots", "url": "https://www.vinayprabhu.com/snapshots", "title": "My virtual abode on these interwebs - Snapshots", "raw_file": "archive/raw/snapshots.html", "text": "Bias is not just in the data foo!\nNetwork Science meets DL arch\nShallow first. Deep next!\nDeterministic initialization of NNs.\nIf saliency cropping is the answer, what is the question?\nLandscape of shortcomings of SIC\nTaxonomy of concerns concerning neural art\nThe eugenicist patriarchs of correlation metrics", "blocks": [ { "tag": "p", "html": "

    Bias is not just in the data foo!

    ", "text": "Bias is not just in the data foo!" }, { "tag": "p", "html": "

    Network Science meets DL arch

    ", "text": "Network Science meets DL arch" }, { "tag": "p", "html": "

    Shallow first. Deep next!

    ", "text": "Shallow first. Deep next!" }, { "tag": "p", "html": "

    Deterministic initialization of NNs. 

    ", "text": "Deterministic initialization of NNs." }, { "tag": "p", "html": "

    If saliency cropping is the answer, what is the question?

    ", "text": "If saliency cropping is the answer, what is the question?" }, { "tag": "p", "html": "

    Landscape of shortcomings of SIC

    ", "text": "Landscape of shortcomings of SIC" }, { "tag": "p", "html": "

    Taxonomy of concerns concerning neural art

    ", "text": "Taxonomy of concerns concerning neural art" }, { "tag": "p", "html": "

    The eugenicist patriarchs of correlation metrics

    ", "text": "The eugenicist patriarchs of correlation metrics" } ], "links": [], "images": [ { "url": "https://lh7-us.googleusercontent.com/sitesv-images-rt/APXyLQeZUHK1IGqHVYPrZf2eAtu15LQoWGBwRN2Pws7XqvjvZH2Lnop4q1P9mVXYJnVz0G_OSX-3JkMyIdBgiQ3OxIhDhJV2smxAMjUqysU8ocrr2n9JeUgXuZrwCVy463OilzNs7wxyRH0zNyGg9joN31QZN49hQ_IBkKw9o61GSjtmor8436MnFheKOg7kFb5OENcaxJFNd43sob-Xke8=w1280", "alt": "" }, { "url": "https://lh7-us.googleusercontent.com/sitesv-images-rt/APXyLQcbbreJ7LoKNs38zG-lfk6WJkhMV4km6t06-rqka60fR6gbplETLwkGGCk5QkuqU4E7BLMOhberAab0sHtDDFnz3WuIMDBS_NodE4R2rChMk20QOsWynf6Vjmkc9sBB6gCdnLB1vURbXXScVyc1GdeSqk5aHDbIagMUIB7Se8__C5jgNx3HrCTn0l5zwyc5J10VSfi1JSFhDaWowwc=w1280", "alt": "" }, { "url": "https://lh7-us.googleusercontent.com/sitesv-images-rt/APXyLQffzJQaZyUgrriVznNgccmQx6geL4kIJSqPiBJIVYMlAx7xg8S1dCF1pftsvOmxKq6V7m5aLqs5Pb0XRtbqUhqDl9s8qBAKLt9JBMhaVZWJt0xK305xxT679KVrtb5RZsMi-EqjWNWWPJ5lhfB6CdWHgat8vfjfmHRZisDApDV_vysHuwt-Tf1aw0Osg84R9w7IXYyWpbCjaoJdjxI=w1280", "alt": "" }, { "url": "https://lh7-us.googleusercontent.com/sitesv-images-rt/APXyLQcVlyMsTJRQZigXDVUK8Qg1gDb0v49Uw2PirwuGhBGJNKeAyxzmsN5mApqcjUxz9oVKQ5SVze1hat6uvFAS8fzsrK5eoKqu4b4wB53X4peO7X6-vxNkr1PwKPsXneZasyUJZ3XzV1Ykel76eGZWE6iN0oPgS3Q2awhTef62Lswx-c2HaD_zWEZlAaVwtuynuYTs_TE-yzElgsVQpVs=w1280", "alt": "" }, { "url": "https://lh7-us.googleusercontent.com/sitesv-images-rt/APXyLQdmxLu6Gr7AdFbUl2XcvlfuiGg1L-Wk15BuP2R_NFJ5l5DHUcxPTH0BynRbMg1efN0RW8IxfzNsumVu9Iyradc8a83SlezTV93pI3WaUMJFJ6nLvmswqwt4yee8TqSweKe5Su0zYuyEdPJF0YyEE4ZpW670UxFm7YpoPCk_at2Ba4xiBNU4PRvraiA-frqYJIK0I9foKSaj04bZLMsbeQv7cSohBk9zVTi_Om_7PRc=w1280", "alt": "" }, { "url": "https://lh7-us.googleusercontent.com/sitesv-images-rt/APXyLQeDElCONCOk1p4pjB-9mlQV63Gj9NFPNwUifo1sQt9KIiIGuOrzFKTz1VLq2-npJxcQrb3x5C1MQrZbrWnGf2qsMdYG437HzSY3Bl7REGCM7njFD_KUUVKbP2t9ncWIyZGFkLFClL3k-Fq7P2eO3AFvT315WEBpHLKbi5vtriXRXBVkDaB83EGER4QgEKoIISQJX32X3GbT9WQKBcVVH_IRulyjVdy8dArvi5KbxRM=w1280", "alt": "" }, { "url": "https://lh7-us.googleusercontent.com/sitesv-images-rt/APXyLQdTovYHTVJk6_0ohQZqFPtn623A248oQFOKINRdU1cAK5SM1xMSHTXrPLFOwroBqhTGO97pCVngWVCy9mta06OBPhsU3Q3TOSRe7uAbQwHc0F-t_pK85t5rzPkpaZPxJAuSygETCHKel3FPNdHnZwGqcqtPqn4p_MNH_2Gkv66O9bfq77EgOYD8dQuEww7PdI5wg2TiWCh3UF_PNb9KrZUtx7-y3qNeRTbgh8dUEPo=w1280", "alt": "" }, { "url": "https://lh7-us.googleusercontent.com/sitesv-images-rt/APXyLQcLSE8ukg6PJgOf4qvWE5hA6Kk2E6QvyW7q5PaM4gOid_ETZVz-1s60D53HOTNQl-4CRNl0qWQ-wwgcEgAvQXCTS0nXLZM70mXe0CnFSM0BZXuvS_txWqaXmIbyL9rzliJjPz9c-7AZdFLP2y7OPc8MLcJ7DUhSfDhD2pyDZvL09DgqB4KApH2H2XpiKTc=w1280", "alt": "" } ], "embeds": [ { "path": "archive/embeds/snapshots-1.html.txt", "text": "Matrix Rain with Blue Neon Text \"Those who ride the tiger can never dismount.\"", "links": [] } ], "iframes": [] }, { "slug": "media", "url": "https://www.vinayprabhu.com/media", "title": "My virtual abode on these interwebs - Media", "raw_file": "archive/raw/media.html", "text": "Mainstream media coverage of some of my research work\nVice (Motherboard)\nVentureBeat awards\nVentureBeat\n(The machine)\nZDNet\nUndark\nIEEE Spectrum\nVox\nWired\nGoogle translation AI botches legal terms 'enjoin,' 'garnish' -research\nThe register\nOn the unethical images  in the Imagenet dataset\nSensor-mag article\nElectronics-for-u ( Make magazine: Indian style )\nVenture-beat coverage\nInterview at\nData-by-the bay\nRSIP Vision mag coverage- Feb- 2018 issue\nArticles and letters published by me in my high school and college years", "blocks": [ { "tag": "h1", "html": "

    Mainstream media coverage of some of my research work

    ", "text": "Mainstream media coverage of some of my research work" }, { "tag": "p", "html": "

    Vice (Motherboard)

    ", "text": "Vice (Motherboard)" }, { "tag": "p", "html": "

    VentureBeat awards

    ", "text": "VentureBeat awards" }, { "tag": "p", "html": "

    VentureBeat
    (The machine)

    ", "text": "VentureBeat (The machine)" }, { "tag": "p", "html": "

    ZDNet

    ", "text": "ZDNet" }, { "tag": "h2", "html": "

    Undark

    ", "text": "Undark" }, { "tag": "h2", "html": "

    IEEE Spectrum

    ", "text": "IEEE Spectrum" }, { "tag": "h2", "html": "

    Vox

    ", "text": "Vox" }, { "tag": "h2", "html": "

    Wired

    ", "text": "Wired" }, { "tag": "h2", "html": "

    The register

    ", "text": "The register" }, { "tag": "p", "html": "

    On the unethical images  in the Imagenet dataset

    ", "text": "On the unethical images  in the Imagenet dataset" }, { "tag": "h2", "html": "

    Sensor-mag article

    ", "text": "Sensor-mag article" }, { "tag": "h2", "html": "

    Electronics-for-u ( Make magazine: Indian style )

    ", "text": "Electronics-for-u ( Make magazine: Indian style )" }, { "tag": "h2", "html": "

    Venture-beat coverage

    ", "text": "Venture-beat coverage" }, { "tag": "p", "html": "

    Interview at Data-by-the bay

    ", "text": "Interview at Data-by-the bay" }, { "tag": "p", "html": "

    RSIP Vision mag coverage- Feb- 2018 issue

    ", "text": "RSIP Vision mag coverage- Feb- 2018 issue" }, { "tag": "h2", "html": "

    Articles and letters published by me in my high school and college years

    ", "text": "Articles and letters published by me in my high school and college years" } ], "links": [ { "text": "Vice (Motherboard)", "url": "https://www.vice.com/en/article/y3gj3v/sexist-ai-is-even-more-sexist-than-we-thought" }, { "text": "VentureBeat awards", "url": "https://venturebeat.com/2020/07/15/announcing-the-ai-innovation-awards-winners-at-transform-2020/" }, { "text": "VentureBeat (The machine)", "url": "https://venturebeat.com/2020/09/20/apparent-racial-bias-found-in-twitter-photo-algorithm/" }, { "text": "ZDNet", "url": "https://www.zdnet.com/article/ethics-of-ai-the-benefits-and-risks-of-artificial-intelligence/" }, { "text": "Undark", "url": "https://undark.org/2021/10/18/computer-scientists-try-to-sidestep-ai-data-dilemma/" }, { "text": "IEEE Spectrum", "url": "https://spectrum.ieee.org/open-ais-powerful-text-generating-tool-is-ready-for-business" }, { "text": "Vox", "url": "https://www.vox.com/videos/2021/3/31/22348722/ai-bias-racial-machine-learning" }, { "text": "Wired", "url": "https://www.wired.com/story/foundations-ai-riddled-errors/" }, { "text": "Google translation AI botches legal terms 'enjoin,' 'garnish' -research", "url": "https://www.google.com/url?q=https%3A%2F%2Fwww.reuters.com%2Farticle%2Fctech-us-alphabet-google-translate-idCAKBN2C612T-OCATC&sa=D&sntz=1&usg=AOvVaw1K1Tdz7Yq_dpHiVThcSH_W" }, { "text": "The register", "url": "https://www.theregister.co.uk/AMP/2019/10/23/ai_dataset_imagenet_consent/?__twitter_impression=true&fbclid=IwAR3WTUzAnyZEzAOEhNbtAun_Z7mxwKBLmNTxKOnHDlDfA56elDWVsHz8TYo" }, { "text": "Sensor-mag article", "url": "https://www.sensorsmag.com/components/smartphone-motion-sensors-healthcare-cusp-a-disruption" }, { "text": "Electronics-for-u ( Make magazine: Indian style )", "url": "https://electronicsforu.com/technology-trends/tech-focus/threat-attacks-smartphone-sensors" }, { "text": "Venture-beat coverage", "url": "http://venturebeat.com/2018/06/22/researchers-trick-watson-ai-into-seeing-cats-as-crazy-quilts-and-cellophane/" }, { "text": "Data-by-the bay", "url": "http://2016.data.bythebay.io/" }, { "text": "RSIP Vision mag coverage- Feb- 2018 issue", "url": "http://www.rsipvision.com/CVPR2017-Saturday/files/assets/common/downloads/Computer%20Vision%20News.pdf" }, { "text": "Articles and letters published by me in my high school and college years", "url": "https://drive.google.com/file/d/1qIlSXApiPLj2tZPFjYXjSvRQ9KNeLtN2/view?usp=sharing" } ], "images": [ { "url": "https://lh7-us.googleusercontent.com/sitesv-images-rt/APXyLQf8n7Hfem6UjywE4TqsPeSK_ul4teTkQKD6KdMuUYlHdCJyhAVCBAN2qw_tyi2LBH24g4Mdhh8HUd_lDrp4kVJ1-gmchmCTx99WeeQUYv6mgdpfx72VVpO0K5nOq1baCzh-GU9Uvf4R_oLozPMy_1peybc1Xv9Uq9hbKCRReIVg733Erv-D_gaaU3CTCCRWOi0_RbS_ONSpVFbUp3rHpGmDzRfCg-DKACscw63mXvs=w1280", "alt": "" }, { "url": "https://lh7-us.googleusercontent.com/sitesv-images-rt/APXyLQeONqa5xm1_qdeVulp0eMUR1qFhTCQyDwsiZOPiVjGiea0BBhP6RFi7uq3tk2sWwdZiksweJNXjcz0e_rMG1PSLWGI5_MCEMu5jls8HdOpoNM8RW3vO4z6_pbSAwWuidt9ROi2vEdDUyKa2Hy67o33Jr-q_CmrSfuQNYljQCtwhPm7lVA0nxXNdhc4fVRWjkT28mt04D7wWqw7LCobtum6eOAT7oihTuIw6jFJNijs=w1280", "alt": "" }, { "url": "https://lh7-us.googleusercontent.com/sitesv-images-rt/APXyLQceIaifsJwvEWis88ChFXgFXKk9OxMdliU--naJ_AAWk21MEuddEd7Yx41DtAlF7LUWDmd_lIY5Rh9fpf0yUzMHhrSgvc688oFKsKhFYENpf6zAT1Q69Zmveq-Fod1gk_PWa1rj7gWg3kE713BgkpJDsR706Wi6sEyWRtgTSagVc4LcPFwCH3nks06d56jeWQqyD93TfPLOWgSBLxkj_kuJUbqUWtRabSx6T--j=w1280", "alt": "" }, { "url": "https://lh7-us.googleusercontent.com/sitesv-images-rt/APXyLQfSjW8dMZ94qFQvjO0Oy0FGH6VmSJa0XSchiWIpLWchwb7VXBUtzrxzW_WmqHmGOIWZcE5eqa0h-qF1NdAgOmRDxy6lTPRWCwCciDcqeEpZgWtI31Wkts2pUCOveQMYC5f0HGL1oDtl_7aqMZjLFTcjg792PhBdQU4QjjayiL0_UaREAgIUV94YYURPsrFRmJX-o5v6N-KrmMZclgSgjpYvszulTqok1GI4PV92EXU=w1280", "alt": "" }, { "url": "https://lh7-us.googleusercontent.com/sitesv-images-rt/APXyLQdeekk5lF7yRT-zmgs8fy360GM3lpOwy-j8VtG7J6DSIUd-nNTiGoX7NJATugcDe-u-bIZRtFbUPkSHFWpWichYweJhaKTME48VBKoc5BtVrZqYBKe_l0XLeBrP241fPj8fRuZZEz6N1XjsnRiaZHzrfGr4759gCqHCnlEFGFJ7XzmHu2xprJwbBfPyY_1q2oL3P8UbXUJYelY3zqjWpEhQR9ahx4kXm1MzCdz17uo=w1280", "alt": "" }, { "url": "https://lh7-us.googleusercontent.com/sitesv-images-rt/APXyLQf2omiISf8EOPNVmOQdRFOHBeNdoAN_NoN1qwS_PGoe_ZEi3cXrDmJk40rLOOVuZkOVaSElreR0hmXtgtDcyhPT4lxsDtDXHn6KFPJ3OR1dfdaXXynSL7kb1pJwrOwFjIajIqa5OS2ms18-J9CpskqPPg9HKg6jjjJsS8mZ70hHRtrB5Lt6U2mJFUDjLAy1GlZe0jI4egNmloYKWKEZ9Ds7CAV_7uvgFQWiXBKswK8=w1280", "alt": "" }, { "url": "https://lh7-us.googleusercontent.com/sitesv-images-rt/APXyLQcZ_awlFHvUUc_A6jRBi-zig3C94ebRnEKV4Vymz7oPrvcVrhapYUlhXNrNBna-WqHMUrOlogDks9MoFKGaUBWN_hNFrzMcKFlyWHNZVijbnZcrOobvP40IrnDx74moO1YwMTTchf6JM1xYmry9BTwf2QN7B2vrF_KpjccduQKyJMipgRMYnna6CQ9Ac7SoIcj3gOXfcQs26qDZUfnlcDUuxqts96EcBIP9jhfg=w1280", "alt": "" }, { "url": "https://lh7-us.googleusercontent.com/sitesv-images-rt/APXyLQfnUiZYlNvH13Lf_DhDV2NF86-kP_1SPX4wHuIlY2hMlchRPA7mo4cII8BRrSo_J8QsWWIMEcbyZjowGn6V19kHNUZy4vxeLRQmCms9MDS_B7BqpBltAYB1bAuUzqyslK0Om5a0IiKESnXchSyHg4aSlYuuyxfBgZfNX4SRkiMkjoXkH1VjlJRwsYMizwLOygEJn_asbRxmIqSI1NxRxkte0ZuVKCbu8X_x4nUcJcU=w1280", "alt": "" }, { "url": "https://lh3.googleusercontent.com/docsubipk/AP9E6xVvaconU2C_Ofxnz72Bewpok5L1VnUfe5tqmlYbt-JoeEtZzuttXTJEtastczMvTh2Q5fzxjmNl7xIYS9mQzybDLVUlih1hoiIOdLSjPCfGNfaul-_ChjdJsIexMzePbQveZuDb5A7sPac49PN6k6eJRcSLUitWsZzIfqGilm8vnguV8EldzaM", "alt": "" }, { "url": "https://lh7-us.googleusercontent.com/sitesv-images-rt/APXyLQcgFfMgACYulXVN192_6cRcx0rBHbs9Y446C3eazjCPfP9I3s_q6ill1cWDa52c0R7COAiyeYBkjaJagRnSuDRqdV6MYySttf9YNZuA67Mvod4Vxh09hzcpVqNqXKZS2-TC8THOVImHlmhLe2kZCy5PJu2HwYJ3lC8ACd9BAVFUsmsvFclu3GZ14rotC98rrZyAhnRYFr0Xf7M-AEE=w1280", "alt": "" }, { "url": "https://lh7-us.googleusercontent.com/sitesv-images-rt/APXyLQd7LQGRXASuir3ZO8Bd-vxqckcbi1wyQnSToro6TyVKwz5N8pZpNMxOK6cUlDzC6zAL3SHrVGDYs8TOxye3-Q3O4cm0iMAMwp7vIpOYKqDe1pUC1P03UKNQ3MiZ8lzx04G6P2FFryfUu0Ev2uZllW5mMncIMFZ5jBOCnOgTjqMG5CGTp_ASdDriXqMuadrD9gYoIFgwbTx12hgNFOc=w1280", "alt": "" }, { "url": "https://lh7-us.googleusercontent.com/sitesv-images-rt/APXyLQctNpVZ8XS1z27Qgl5wZnjbfzZLQdBHfzVPO3d0Pplj5Y4ubzZDWlN47oslQCpjKvqUbd4OfvHz57WVHXQdccE-MhREOBd9eiC5mab7X1ZYlptX_5Msdmb7nBN2PEC6cLV_Rr4DpgMuau9GryhvkaxGO_YQV0awzvLs-bFlQ1nv9X1DRQLWqqGCG0RcAe-3_DZ-O1oIJn2cjVkSYHs=w1280", "alt": "" }, { "url": "https://lh7-us.googleusercontent.com/sitesv-images-rt/APXyLQcyT6qTA_xshDtcFDsKKXLO-nOmKeR4xDJ3nVc0UvjbN8ecSnM1M3kFbH-mZ5YmYDu3vlrv-FFvkUG8rUbTxOfjMHsQpB3tUg8-RyZo3n_TFqwnPN3lWjRIFDUSm6pwU462Xqe4rUtoY-OiI6M80dDOQzJu960stDpQzN9vMeVLdw2v7EO7QdkZRK1nJ8z_z0UkP9LhlbtoqYursLE=w1280", "alt": "" }, { "url": "https://lh7-us.googleusercontent.com/sitesv-images-rt/APXyLQcrV8h_gCo3BZOOmGDVs52dEtcihGlXuraBDr-EYusXpChHjkvsT6CWdFVnc98m1Q1PfDctHfTOspEPIqDmhhCHivX6iyPaZLHd3cswlSVFWND2HZtGSoGylMBqpHREuES4hFIcpQwec3FxgL7k31-Sh6JKy6LTUjbp-4qelayVrENSIoJx8Wd6QMhNEhMWMYwgAKuw283W0ZvgTHg=w1280", "alt": "" }, { "url": "https://lh7-us.googleusercontent.com/sitesv-images-rt/APXyLQf8PiIUD3foTh4VVP_cpVmM48Y7NpV4-0tM_vehtfZm_1THEBWoVgflhBIH47siXqGwmTze8imheRYnPPRT09xkQ2iHUDkCtbsxlk6DK0_ApLxZN29y_PzOoMIC2MnESF93adUCIaT0ZceWn21EoygwYsrOrW9J6aFdMeuUlU3_nCALLPaPk2iX0OZxM3R09BfUUwvZBUL0tmh0=w1280", "alt": "" } ], "embeds": [ { "path": "archive/embeds/media-1.html.txt", "text": "Matrix Rain with Blue Neon Text \"Those who ride the tiger can never dismount.\"", "links": [] } ], "iframes": [ { "jsname": "L5Fo6c", "class": [ "YMEQtf" ], "sandbox": [ "allow-scripts", "allow-popups", "allow-forms", "allow-same-origin", "allow-popups-to-escape-sandbox", "allow-downloads", "allow-modals", "allow-storage-access-by-user-activation" ], "frameborder": "0", "aria-label": "YouTube Video, DBTB INT Vinay Prabhu r", "src": "https://www.youtube.com/embed/wV0dDIoppsw?embed_config=%7B%22enc%22:%22AXH1ezmFJm4JaYgahaz0VV1nVvDn3xG2MHX53buZPSz-K-GpXWRw4RruwzmTpNUyPvm9xCQhI3UPa8MUW99I7rv6Cvfks85t8TtuxBce3d2QrtkMyAZpz1LIzof08_Ad8gDQLYBibCA5dw6im1aqTNCWP5IQ0diAFTRNm_rBLDUd0PpO%22%7D&enablejsapi=1&errorlinks=1&origin=https://www.vinayprabhu.com&vl=1", "allowfullscreen": "" } ] }, { "slug": "miscellaneous", "url": "https://www.vinayprabhu.com/miscellaneous", "title": "My virtual abode on these interwebs - Miscellaneous", "raw_file": "archive/raw/miscellaneous.html", "text": "Miscellaneous\nOde to the inner cricket-tragic.\nClubs represented: Oporto Cricket & Lawn Tennis Club and Northhills Knights CC, Pittsburgh, PA\nOde to the inner cricket-tragic\n.\nClubs represented:\nOporto Cricket & Lawn Tennis Club\nand  Northhills Knights CC, Pittsburgh, PA\nCareer stats:\ncricclubs.com/PittsburghCA/viewPlayer.do?playerId=655328&clubId=4041\nCamp BYTE: Burners for Ÿntelligent Technology Emancipation\nhttps://www.indiegogo.com/projects/b-y-t-e#/", "blocks": [ { "tag": "h1", "html": "

    Miscellaneous

    ", "text": "Miscellaneous" }, { "tag": "h2", "html": "

    Ode to the inner cricket-tragic. 

    ", "text": "Ode to the inner cricket-tragic ." }, { "tag": "h2", "html": "

    Clubs represented: Oporto Cricket & Lawn Tennis Club and  Northhills Knights CC, Pittsburgh, PA

    ", "text": "Clubs represented: Oporto Cricket & Lawn Tennis Club and  Northhills Knights CC, Pittsburgh, PA" }, { "tag": "p", "html": "

    Career stats: cricclubs.com/PittsburghCA/viewPlayer.do?playerId=655328&clubId=4041

    ", "text": "Career stats: cricclubs.com/PittsburghCA/viewPlayer.do?playerId=655328&clubId=4041" }, { "tag": "p", "html": "

    Camp BYTE: Burners for Ÿntelligent Technology Emancipation 

    ", "text": "Camp BYTE: Burners for Ÿntelligent Technology Emancipation" }, { "tag": "p", "html": "

    https://www.indiegogo.com/projects/b-y-t-e#/ 

    ", "text": "https://www.indiegogo.com/projects/b-y-t-e#/" } ], "links": [ { "text": "Ode to the inner cricket-tragic.", "url": "https://www.vinayprabhu.com/miscellaneous#h.p_BUgVkpzOglNA" }, { "text": "Clubs represented: Oporto Cricket & Lawn Tennis Club and Northhills Knights CC, Pittsburgh, PA", "url": "https://www.vinayprabhu.com/miscellaneous#h.p_VL2PWoehglp-" }, { "text": "Ode to the inner cricket-tragic", "url": "https://youtu.be/7TVL-gyBILM" }, { "text": "Oporto Cricket & Lawn Tennis Club", "url": "https://www.oportocricketclub.com/" }, { "text": "cricclubs.com/PittsburghCA/viewPlayer.do?playerId=655328&clubId=4041", "url": "https://cricclubs.com/PittsburghCA/viewPlayer.do?playerId=655328&clubId=4041" }, { "text": "https://www.indiegogo.com/projects/b-y-t-e#/", "url": "https://www.indiegogo.com/projects/b-y-t-e#/" } ], "images": [ { "url": "https://lh7-us.googleusercontent.com/sitesv-images-rt/AMxu72ufa4Zp8Ru5pFkme5yxzeQtCvNhrkaW6HYlgizSCkLGbwSwtVw0AKQqM6iaCNvxLAOgoXzemzxlZ3EyP1EcxVS2CqHa4QrK5ZUm_uykUANYwyhMFKQFXRset89oO6AUCHb9deAj2hvqAL6iYZwH7O3L79mqG7TTG69nrmVBDM43vpWbuZNrIrG-qv40z1UV0t5FSBLFd7D7Vo5guyw=w1280", "alt": "" }, { "url": "https://lh7-us.googleusercontent.com/sitesv-images-rt/AMxu72uuOujrpAH1JTPaEKjZC138aKIkt96d3hHQZqRqgsEjk9QgXYEZ1ZqACdF1MJis4KWZN5sC-PJK9s5us1kqfRm2jlfIs_w5mJRKvwKez19X0l6Gy9Ib-vJfErLRPosSMCy4ifBOhI2TywI90UXFXniR-Q8MdRN4FkZSO-PCItab2Mb1k9VbA8oqrL_w6z6kXF4oI2DHEDVIK2Hf5fA=w1280", "alt": "" } ], "embeds": [ { "path": "archive/embeds/miscellaneous-1.html.txt", "text": "Matrix Rain with Blue Neon Text \"Those who ride the tiger can never dismount.\"", "links": [] } ], "iframes": [] }, { "slug": "spice-cabinet", "url": "https://www.vinayprabhu.com/spice-cabinet", "title": "My virtual abode on these interwebs - SPICE-cabinet", "raw_file": "archive/raw/spice-cabinet.html", "text": "Oodles of ODDs: On out-of-distribution histrionics in deep learning\nX-formers architecture survey\nX-is all you need\nPapers are hard to write. Survey papers are just that much harder. From the authors' perspective, challenges include the responsibility to not erase out important work being done by (sometimes) adversarially aligned research groups, finding the right semantic clustering to sub-categorize individual contributions, controlling for the verbosity and length of the final paper, ensuring an optimal mixing of personal opinion and the innate narratives in the paper(s) being cited, version controlling, ease of updating, and also the aesthetics of presentation. From the reader's viewpoint, challenges include ease of reading, single-snapshot summarizability, portability, and being given the agency to edit or fork their own copies.\nTaking cues from the emergence of the cheat-sheet culture in machine learning and the virtues of living editable documentation and version control, we propose an interactive and live SVG format based methodology that we term SPICE: Survey Papers as Interactive Cheat-sheet Embedding. We cover the technical details behind constructing SPICEs and present an example gallery covering `hot button' areas in machine learning such as Out of distribution detection, the `All you need' histrionics and Transformer architectures.\n🖥️\nGithub:\nhttps://github.com/vinayprabhu/SPICEs/blob/main/README.md\n📹\nYoutube:\nhttps://youtu.be/uzx2gxOhn14", "blocks": [ { "tag": "p", "html": "

    Oodles of ODDs: On out-of-distribution histrionics in deep learning

    ", "text": "Oodles of ODDs: On out-of-distribution histrionics in deep learning" }, { "tag": "p", "html": "

    X-formers architecture survey

    ", "text": "X-formers architecture survey" }, { "tag": "p", "html": "

    X-is all you need

    ", "text": "X-is all you need" }, { "tag": "p", "html": "

    Papers are hard to write. Survey papers are just that much harder. From the authors' perspective, challenges include the responsibility to not erase out important work being done by (sometimes) adversarially aligned research groups, finding the right semantic clustering to sub-categorize individual contributions, controlling for the verbosity and length of the final paper, ensuring an optimal mixing of personal opinion and the innate narratives in the paper(s) being cited, version controlling, ease of updating, and also the aesthetics of presentation. From the reader's viewpoint, challenges include ease of reading, single-snapshot summarizability, portability, and being given the agency to edit or fork their own copies.

    ", "text": "Papers are hard to write. Survey papers are just that much harder. From the authors' perspective, challenges include the responsibility to not erase out important work being done by (sometimes) adversarially aligned research groups, finding the right semantic clustering to sub-categorize individual contributions, controlling for the verbosity and length of the final paper, ensuring an optimal mixing of personal opinion and the innate narratives in the paper(s) being cited, version controlling, ease of updating, and also the aesthetics of presentation. From the reader's viewpoint, challenges include ease of reading, single-snapshot summarizability, portability, and being given the agency to edit or fork their own copies." }, { "tag": "p", "html": "

    Taking cues from the emergence of the cheat-sheet culture in machine learning and the virtues of living editable documentation and version control, we propose an interactive and live SVG format based methodology that we term SPICE: Survey Papers as Interactive Cheat-sheet Embedding. We cover the technical details behind constructing SPICEs and present an example gallery covering `hot button' areas in machine learning such as Out of distribution detection, the `All you need' histrionics and Transformer architectures.

    ", "text": "Taking cues from the emergence of the cheat-sheet culture in machine learning and the virtues of living editable documentation and version control, we propose an interactive and live SVG format based methodology that we term SPICE: Survey Papers as Interactive Cheat-sheet Embedding. We cover the technical details behind constructing SPICEs and present an example gallery covering `hot button' areas in machine learning such as Out of distribution detection, the `All you need' histrionics and Transformer architectures." }, { "tag": "p", "html": "

    🖥️ Github: https://github.com/vinayprabhu/SPICEs/blob/main/README.md

    ", "text": "🖥️ Github: https://github.com/vinayprabhu/SPICEs/blob/main/README.md" }, { "tag": "p", "html": "

    📹Youtube: https://youtu.be/uzx2gxOhn14

    ", "text": "📹 Youtube: https://youtu.be/uzx2gxOhn14" } ], "links": [ { "text": "Oodles of ODDs: On out-of-distribution histrionics in deep learning", "url": "https://www.vinayprabhu.com/spice-cabinet/ood" }, { "text": "X-formers architecture survey", "url": "https://www.vinayprabhu.com/spice-cabinet/transformers" }, { "text": "X-is all you need", "url": "https://www.vinayprabhu.com/spice-cabinet/all-you-need" }, { "text": "🖥️", "url": "https://emojipedia.org/desktop-computer/" }, { "text": "https://github.com/vinayprabhu/SPICEs/blob/main/README.md", "url": "https://github.com/vinayprabhu/SPICEs/blob/main/README.md" }, { "text": "https://youtu.be/uzx2gxOhn14", "url": "https://youtu.be/uzx2gxOhn14" } ], "images": [ { "url": "https://lh7-us.googleusercontent.com/sitesv-images-rt/APXyLQcVX0cAfYNyl0Wexx1uk7KwD43XMT2TyxJB9WTp0-oB5hBaN9u4DULrOwuPhDQUEVr3dwzXCaIcZtC3D_rzDXn9mtvRUM2yiQSSmYCuk0DgShFimBDOQUtGvG2Bbfu5RP37sEvH9tga3SxkDGknzzCdxbIStRP2cJ5IZ1aqKgqhYSRd0fFPFdCx9IOtf7rIj-mInBDI_BHIMF3MP4e9IbnhU6yY5DbKmtayDydKAR0=w1280", "alt": "" }, { "url": "https://lh7-us.googleusercontent.com/sitesv-images-rt/APXyLQdjSmRmC8--cLj7uZusAZgreCQUSZsejenxGXSStY1PPh4licBM-ed_xfbOJWD48jv3JD2uz2m7E6AGNEJMbwEnLXGH2LMRztdA6cjhGHkz8NK0zC8wL6VvR4K-Yt8OMi-O3fS6v4iGpJF8xohSx_vJNeSvO8L8EoocTUjZhPJy6bTI59djdV11locAma8p5BJHH4ve-Q517GmcLzkdHzxxOUqZQlNjWtcNgOGhkO4=w1280", "alt": "" }, { "url": "https://lh7-us.googleusercontent.com/sitesv-images-rt/APXyLQekqEoLP-t7RMLb2_a9bGJOq439Fj7-Yq9aymOuc-N6PmT6_t2F-TujqDJRvqwlqSJ-3Fu7l2cPXiNdEv0F6FVzjrUTsc25aO695FM5ftXFWy_WNMbD1YKs827t1rLCN3JnDVGf9KhWFsx2TlRM5mmvsM9o5AvaWLoAAb0CiJ_UlL6XVwVOS0GNPf32JYjyQ_7zEUqWrfc9Ss5LKqPainY_Evfpt1I8tSYnOk6l=w1280", "alt": "" } ], "embeds": [ { "path": "archive/embeds/spice-cabinet-1.html.txt", "text": "Matrix Rain with Blue Neon Text \"Those who ride the tiger can never dismount.\"", "links": [] } ], "iframes": [] }, { "slug": "spice-cabinet/ood", "url": "https://www.vinayprabhu.com/spice-cabinet/ood", "title": "My virtual abode on these interwebs - OOD", "raw_file": "archive/raw/spice-cabinet--ood.html", "text": "OODles of ODDs: Surveying the landscape of Out-of-distribution work in recent ML literature\nIn this work, we chart out the landscape of ideas encapsulated within the ambit of Out-of-Distribution (OOD) or Out-of-distribution-detection (ODD) labels in ML literature\n.\nThe SPICE for this is disseminated via two formats.\nStatic SPICE\n(PDF with URLs to the papers)\nInteractive SPICE below. (Click on the links and the representative Image appears)\nOur wish list for the field\n1.\nNSL\n(Neural Structured Learning):\nand it’s ramifications on OOD still remains to be explored:\nDid you just harness NSL to get a better top-k accuracy on your precious little validation set? Check the newly minted model's OOD susceptibility levels. You\nmight\nbe in for a surprise.\n2.\nEntangled domains:\nIf you, as a machine learner have ever worked on User-authentication (Not\nuser-surveillance\n. But anonymized\nuser-authentication\n), then the [ User | Not-user ] classification is IMHO the ultimate stress-test (perhaps alongside new species identification) for OOD algorithms. In this case, there is zero hand-waviness as to what In-distribution and Out-of-distribution. In-distribution captures all the vagaries captured the user-enrollment. The out-of-distribution captures everything in the universe that is NOT generated by the User you are trying to authenticate. Try your fancy-schmancy, ahem, \"ODD algorithms\" from your Computer-vision biased SOTA and an attitude-correct and a reality check is in store for you!\n3. X-MNISTs are far harder than you think\n:\nMost\nif y'all think of digit-glyph classification when you think of MNIST/QMNIST/Kannada-MNIST. However these datasets also carry the volunteer-ID information associated with the digits. (Example See\nthis\ntensor for Kannada-MNIST). Should you  split these datasets to be in-cohort volunteers authoring the digits and pose ODD as out-of-cohort-volunteer detection, your easy-little MNIST dataset turns into an entangled mess in a jiffy. Give it a try!\n4. Fine-grained classification:\nWhile acknowledging the need for standardized comparisons, we remark that the OOD landscape is littered with strange matchups drawn from Tiny-ImageNet , LSUN, CIFAR-x and SVHN datasets that seem rather ad-hoc. This coincides with this uncanny side-stepping of the world of\nfine-grained classification datasets\n.\nDon't they constitute the Rosetta stone for the field?\n5. Anomaly detection / outlier detection / novelty detection\n:[ **Insert  Distracted Boyfriend meme here** ]. In the pre-deep learning Neanderthal era, loads of interesting and useful ideas were\n6. Mind the Domain-bias:\nThis should seem like a no-brainer, but I'd reckon its worth reiterating that there's exists a vast world outside of Machine Learning outside of Computer Vision. The intuitions developed on a bunch of idiosyncratic datasets in Computer Vision do not / cannot and will not hold true when you switch the application. As someone who grapples with human kinematics and motion-sensor tensors on a daily basis, I can vouch that  when it comes to transfer learning, optimal architecture design , learning-rate regimes and ODD, most best-practices developed in Computer Vision do not hold true!\n7. Hypothesis testing literature:\nSeriously people. There's an entire rich sub-domain of stats called Hypothesis testing that the DL-bros have totally side-stepped. Take a look. Especially at this\nHyppo\n8. Don’t ignore calibration\n:\nI've observed that often model calibration ideas are weirdly  kept separated from the OOD question. Isn't efficient ODD at the heart of a well calibrated model?\n9. Mind the loopholes in falling back on humans\n:\nFace-based surveillance systems deployed in\nreal world\nare, in fact, human-in-the-loop systems. yet they falter. This should not come as a surprise. The human vision system is NOT bias-free! Mind the 3 Ps: Psychonomics, Prosopagnosia, Pareidolia.\n10. In defense of the NN- Isolation bias:\nMuch akin to Model calibration, the sub-domains of\nMembership inference attacks, Dataset distillation, Lottery-ticket-hypothesis-type model compression and data-augmentation techniques strongly influence OOD susceptibility. In the case of\n11. Emergent lure of Africa as an OOD proving ground\n: As much of the imagined occident and the Far East falls under the sway of Big Data, it gets accommodated as ‘in-distribution’. This is quickly eliciting a predictivism driven ‘observer effect’ where Algorithms trained in familiar “in-distributed” settings  excelling on in-distributed data are ‘Meh!’. This sets the stage for an emergent lure of Africa as an OOD proving ground where algorithms trained in in-distributed occidental bastions endure their baptism by fire.", "blocks": [ { "tag": "h1", "html": "

    OODles of ODDs: Surveying the landscape of Out-of-distribution work in recent ML literature

    ", "text": "OODles of ODDs: Surveying the landscape of Out-of-distribution work in recent ML literature" }, { "tag": "p", "html": "

    In this work, we chart out the landscape of ideas encapsulated within the ambit of Out-of-Distribution (OOD) or Out-of-distribution-detection (ODD) labels in ML literature.
    The SPICE for this is disseminated via two formats.

    ", "text": "In this work, we chart out the landscape of ideas encapsulated within the ambit of Out-of-Distribution (OOD) or Out-of-distribution-detection (ODD) labels in ML literature . The SPICE for this is disseminated via two formats." }, { "tag": "li", "html": "

  • Static SPICE (PDF with URLs to the papers)

    • ", "text": "Static SPICE (PDF with URLs to the papers)" }, { "tag": "li", "html": "

  • Interactive SPICE below. (Click on the links and the representative Image appears)

    • ", "text": "Interactive SPICE below. (Click on the links and the representative Image appears)" }, { "tag": "p", "html": "

    Our wish list for the field

    ", "text": "Our wish list for the field" }, { "tag": "p", "html": "

    1.NSL (Neural Structured Learning): and it’s ramifications on OOD still remains to be explored: Did you just harness NSL to get a better top-k accuracy on your precious little validation set? Check the newly minted model's OOD susceptibility levels. You might be in for a surprise.

    ", "text": "1. NSL (Neural Structured Learning): and it’s ramifications on OOD still remains to be explored: Did you just harness NSL to get a better top-k accuracy on your precious little validation set? Check the newly minted model's OOD susceptibility levels. You might be in for a surprise." }, { "tag": "p", "html": "

    2. Entangled domains: If you, as a machine learner have ever worked on User-authentication (Not user-surveillance. But anonymized user-authentication ), then the [ User | Not-user ] classification is IMHO the ultimate stress-test (perhaps alongside new species identification) for OOD algorithms. In this case, there is zero hand-waviness as to what In-distribution and Out-of-distribution. In-distribution captures all the vagaries captured the user-enrollment. The out-of-distribution captures everything in the universe that is NOT generated by the User you are trying to authenticate. Try your fancy-schmancy, ahem, \"ODD algorithms\" from your Computer-vision biased SOTA and an attitude-correct and a reality check is in store for you!

    ", "text": "2. Entangled domains: If you, as a machine learner have ever worked on User-authentication (Not user-surveillance . But anonymized user-authentication ), then the [ User | Not-user ] classification is IMHO the ultimate stress-test (perhaps alongside new species identification) for OOD algorithms. In this case, there is zero hand-waviness as to what In-distribution and Out-of-distribution. In-distribution captures all the vagaries captured the user-enrollment. The out-of-distribution captures everything in the universe that is NOT generated by the User you are trying to authenticate. Try your fancy-schmancy, ahem, \"ODD algorithms\" from your Computer-vision biased SOTA and an attitude-correct and a reality check is in store for you!" }, { "tag": "p", "html": "

    3. X-MNISTs are far harder than you think: Most  if y'all think of digit-glyph classification when you think of MNIST/QMNIST/Kannada-MNIST. However these datasets also carry the volunteer-ID information associated with the digits. (Example See this tensor for Kannada-MNIST). Should you  split these datasets to be in-cohort volunteers authoring the digits and pose ODD as out-of-cohort-volunteer detection, your easy-little MNIST dataset turns into an entangled mess in a jiffy. Give it a try!

    ", "text": "3. X-MNISTs are far harder than you think : Most if y'all think of digit-glyph classification when you think of MNIST/QMNIST/Kannada-MNIST. However these datasets also carry the volunteer-ID information associated with the digits. (Example See this tensor for Kannada-MNIST). Should you  split these datasets to be in-cohort volunteers authoring the digits and pose ODD as out-of-cohort-volunteer detection, your easy-little MNIST dataset turns into an entangled mess in a jiffy. Give it a try!" }, { "tag": "p", "html": "

    4. Fine-grained classification: While acknowledging the need for standardized comparisons, we remark that the OOD landscape is littered with strange matchups drawn from Tiny-ImageNet , LSUN, CIFAR-x and SVHN datasets that seem rather ad-hoc. This coincides with this uncanny side-stepping of the world of  fine-grained classification datasets. Don't they constitute the Rosetta stone for the field? 

    ", "text": "4. Fine-grained classification: While acknowledging the need for standardized comparisons, we remark that the OOD landscape is littered with strange matchups drawn from Tiny-ImageNet , LSUN, CIFAR-x and SVHN datasets that seem rather ad-hoc. This coincides with this uncanny side-stepping of the world of fine-grained classification datasets . Don't they constitute the Rosetta stone for the field?" }, { "tag": "p", "html": "

    5. Anomaly detection / outlier detection / novelty detection:[ **Insert  Distracted Boyfriend meme here** ]. In the pre-deep learning Neanderthal era, loads of interesting and useful ideas were 

    ", "text": "5. Anomaly detection / outlier detection / novelty detection :[ **Insert  Distracted Boyfriend meme here** ]. In the pre-deep learning Neanderthal era, loads of interesting and useful ideas were" }, { "tag": "p", "html": "

    6. Mind the Domain-bias:  This should seem like a no-brainer, but I'd reckon its worth reiterating that there's exists a vast world outside of Machine Learning outside of Computer Vision. The intuitions developed on a bunch of idiosyncratic datasets in Computer Vision do not / cannot and will not hold true when you switch the application. As someone who grapples with human kinematics and motion-sensor tensors on a daily basis, I can vouch that  when it comes to transfer learning, optimal architecture design , learning-rate regimes and ODD, most best-practices developed in Computer Vision do not hold true!

    ", "text": "6. Mind the Domain-bias: This should seem like a no-brainer, but I'd reckon its worth reiterating that there's exists a vast world outside of Machine Learning outside of Computer Vision. The intuitions developed on a bunch of idiosyncratic datasets in Computer Vision do not / cannot and will not hold true when you switch the application. As someone who grapples with human kinematics and motion-sensor tensors on a daily basis, I can vouch that  when it comes to transfer learning, optimal architecture design , learning-rate regimes and ODD, most best-practices developed in Computer Vision do not hold true!" }, { "tag": "p", "html": "

    7. Hypothesis testing literature: Seriously people. There's an entire rich sub-domain of stats called Hypothesis testing that the DL-bros have totally side-stepped. Take a look. Especially at this Hyppo

    ", "text": "7. Hypothesis testing literature: Seriously people. There's an entire rich sub-domain of stats called Hypothesis testing that the DL-bros have totally side-stepped. Take a look. Especially at this Hyppo" }, { "tag": "p", "html": "

    8. Don’t ignore calibration : I've observed that often model calibration ideas are weirdly  kept separated from the OOD question. Isn't efficient ODD at the heart of a well calibrated model?

    ", "text": "8. Don’t ignore calibration : I've observed that often model calibration ideas are weirdly  kept separated from the OOD question. Isn't efficient ODD at the heart of a well calibrated model?" }, { "tag": "p", "html": "

    9. Mind the loopholes in falling back on humans : Face-based surveillance systems deployed in real world are, in fact, human-in-the-loop systems. yet they falter. This should not come as a surprise. The human vision system is NOT bias-free! Mind the 3 Ps: Psychonomics, Prosopagnosia, Pareidolia.

    ", "text": "9. Mind the loopholes in falling back on humans : Face-based surveillance systems deployed in real world are, in fact, human-in-the-loop systems. yet they falter. This should not come as a surprise. The human vision system is NOT bias-free! Mind the 3 Ps: Psychonomics, Prosopagnosia, Pareidolia." }, { "tag": "p", "html": "

    10. In defense of the NN- Isolation bias: Much akin to Model calibration, the sub-domains of Membership inference attacks, Dataset distillation, Lottery-ticket-hypothesis-type model compression and data-augmentation techniques strongly influence OOD susceptibility. In the case of 

    ", "text": "10. In defense of the NN- Isolation bias: Much akin to Model calibration, the sub-domains of Membership inference attacks, Dataset distillation, Lottery-ticket-hypothesis-type model compression and data-augmentation techniques strongly influence OOD susceptibility. In the case of" }, { "tag": "p", "html": "

    11. Emergent lure of Africa as an OOD proving ground: As much of the imagined occident and the Far East falls under the sway of Big Data, it gets accommodated as ‘in-distribution’. This is quickly eliciting a predictivism driven ‘observer effect’ where Algorithms trained in familiar “in-distributed” settings  excelling on in-distributed data are ‘Meh!’. This sets the stage for an emergent lure of Africa as an OOD proving ground where algorithms trained in in-distributed occidental bastions endure their baptism by fire.

    ", "text": "11. Emergent lure of Africa as an OOD proving ground : As much of the imagined occident and the Far East falls under the sway of Big Data, it gets accommodated as ‘in-distribution’. This is quickly eliciting a predictivism driven ‘observer effect’ where Algorithms trained in familiar “in-distributed” settings  excelling on in-distributed data are ‘Meh!’. This sets the stage for an emergent lure of Africa as an OOD proving ground where algorithms trained in in-distributed occidental bastions endure their baptism by fire." } ], "links": [ { "text": "Static SPICE", "url": "https://github.com/vinayprabhu/OODles_of_ODDs/raw/main/OOD_SPICE.pdf" }, { "text": "NSL", "url": "https://www.tensorflow.org/neural_structured_learning" }, { "text": "this", "url": "https://github.com/vinayprabhu/Kannada_MNIST/blob/master/data/output_tensors/y_Ka_mnist_userwise.npy" }, { "text": "fine-grained classification datasets", "url": "https://paperswithcode.com/task/fine-grained-image-classification" }, { "text": "Hyppo", "url": "https://hyppo.neurodata.io/" }, { "text": "real world", "url": "https://arstechnica.com/tech-policy/2020/06/detroit-police-chief-admits-facial-recognition-is-wrong-96-of-the-time/#:~:text=Detroit%27s%20police%20chief%20admitted%20on,match%20from%20facial%20recognition%20software." } ], "images": [ { "url": "https://lh7-us.googleusercontent.com/sitesv-images-rt/APXyLQdy9Zw4XJCUgBuJ366sG_IhMoZ6KTqBukfS4sa0NT83iVAlrPD8gOcnoUZLew1U30BlhAs4qAW6EHgGYni-THMz9rfhrXDKSyC_fhanm2bUzklZa7xcFihlngmBO46u79ePGTBorLwSRJ-bd96lKbq73yV0QDB6wQUzII4l7h9bxUc3S5LZrSGk_nAkT_YSC5lkq9NJkKCBB7Lmy6Q4ocHcaBwV0o82dMbwIJUjcBM=w1280", "alt": "" } ], "embeds": [ { "path": "archive/embeds/spice-cabinet--ood-1.html.txt", "text": "Matrix-Themed Flipsnack Widget Fullscreen", "links": [ { "text": "Fullscreen", "url": "https://cdn.flipsnack.com/widget/v2/widget.html?hash=vp84kcde47" } ] }, { "path": "archive/embeds/spice-cabinet--ood-2.html.txt", "text": "Matrix Rain with Blue Neon Text \"Those who ride the tiger can never dismount.\"", "links": [] } ], "iframes": [ { "jsname": "WMhH6e", "class": [ "YMEQtf" ], "frameborder": "0", "sandbox": [ "allow-scripts", "allow-popups", "allow-forms", "allow-same-origin", "allow-popups-to-escape-sandbox", "allow-downloads", "allow-storage-access-by-user-activation" ], "id": "1301e0c1e7d58073_1", "name": "1301e0c1e7d58073_1", "scrolling": "no", "title": "Custom embed", "aria-label": "Custom embed" } ] }, { "slug": "spice-cabinet/transformers", "url": "https://www.vinayprabhu.com/spice-cabinet/transformers", "title": "My virtual abode on these interwebs - Transformers", "raw_file": "archive/raw/spice-cabinet--transformers.html", "text": "SPICE of Efficient Transformers: A Survey by\nTay et al\nTL-DR:\nSurveying the\nX\nin\nX\n-formers\nThe SPICE for this is disseminated via two formats.\nStatic SPICE\n(PDF with URLs to the papers)\nInteractive SPICE below. (Click on the links and the representative Image appears. Click on 'Fullscreen')", "blocks": [ { "tag": "h1", "html": "

    SPICE of Efficient Transformers: A Survey by Tay et al

    ", "text": "SPICE of Efficient Transformers: A Survey by Tay et al" }, { "tag": "p", "html": "

    TL-DR:
    Surveying the X in X-formers

    ", "text": "TL-DR: Surveying the X in X -formers" }, { "tag": "p", "html": "

    The SPICE for this is disseminated via two formats.

    ", "text": "The SPICE for this is disseminated via two formats." }, { "tag": "li", "html": "

  • Static SPICE (PDF with URLs to the papers)

    • ", "text": "Static SPICE (PDF with URLs to the papers)" }, { "tag": "li", "html": "

  • Interactive SPICE below. (Click on the links and the representative Image appears. Click on 'Fullscreen')

    • ", "text": "Interactive SPICE below. (Click on the links and the representative Image appears. Click on 'Fullscreen')" } ], "links": [ { "text": "Tay et al", "url": "https://arxiv.org/abs/2009.06732" }, { "text": "Static SPICE", "url": "https://github.com/vinayprabhu/OODles_of_ODDs/raw/main/SPICE_CABINET/efficient_transformers_spice.pdf" } ], "images": [ { "url": "https://lh7-us.googleusercontent.com/sitesv-images-rt/APXyLQcGg9_7yVENI-l6_G9kBlVfdQ7Nn7CcQVEzKhYjSLi6FpIMRZNaw75r8OMWGKGsM8eVu2jskcNwIsQFXPTGiFeuFIlRugHuTsO4ej88ZdOKEEWbgjJDLzGQWimnncvnd4mgLp3NnMF92znHCn6N5bc_AxZzN_MvKuUxqwUX9uxgZBpRxQgDIZTYzhqbds07T0qfS_Osb9nPVL2bxzPnkQPiaS0wtkU5f-FIw2GAHq0=w1280", "alt": "" } ], "embeds": [ { "path": "archive/embeds/spice-cabinet--transformers-1.html.txt", "text": "Fullscreen", "links": [ { "text": "Fullscreen", "url": "https://cdn.flipsnack.com/widget/v2/widget.html?hash=c8wkye5hp8" } ] }, { "path": "archive/embeds/spice-cabinet--transformers-2.html.txt", "text": "Matrix Rain with Blue Neon Text \"Those who ride the tiger can never dismount.\"", "links": [] } ], "iframes": [ { "jsname": "WMhH6e", "class": [ "YMEQtf" ], "frameborder": "0", "sandbox": [ "allow-scripts", "allow-popups", "allow-forms", "allow-same-origin", "allow-popups-to-escape-sandbox", "allow-downloads", "allow-storage-access-by-user-activation" ], "id": "7daffa09be0fea86_1", "name": "7daffa09be0fea86_1", "scrolling": "no", "title": "Custom embed", "aria-label": "Custom embed" } ] }, { "slug": "spice-cabinet/all-you-need", "url": "https://www.vinayprabhu.com/spice-cabinet/all-you-need", "title": "My virtual abode on these interwebs - All-you-need", "raw_file": "archive/raw/spice-cabinet--all-you-need.html", "text": "X- is all you need!\nThe success enjoyed by the\nAttention is all you need\nspurred an avalanche of papers whose title or content banked off of the X-\nis all you need\nquip 😅. In\nthe SPICE here\n, we present a survey of all these papers that allows the reader to summarize where and why these quips occur.\nDataset: Here's the\nGithub repo\nPDF:\nX-is-all-you-need SPICE", "blocks": [ { "tag": "h1", "html": "

    X- is all you need!

    ", "text": "X- is all you need!" }, { "tag": "p", "html": "

    The success enjoyed by the Attention is all you need spurred an avalanche of papers whose title or content banked off of the X-is all you need quip 😅. In the SPICE here, we present a survey of all these papers that allows the reader to summarize where and why these quips occur.

    ", "text": "The success enjoyed by the Attention is all you need spurred an avalanche of papers whose title or content banked off of the X- is all you need quip 😅. In the SPICE here , we present a survey of all these papers that allows the reader to summarize where and why these quips occur." }, { "tag": "li", "html": "

  • Dataset: Here's the Github repo

    • ", "text": "Dataset: Here's the Github repo" }, { "tag": "li", "html": "

  • PDF: X-is-all-you-need SPICE 

    • ", "text": "PDF: X-is-all-you-need SPICE" } ], "links": [ { "text": "Attention is all you need", "url": "https://arxiv.org/abs/1706.03762" }, { "text": "Github repo", "url": "https://github.com/vinayprabhu/X-is-all-you-need" }, { "text": "X-is-all-you-need SPICE", "url": "https://github.com/vinayprabhu/SPICEs/raw/main/SPICE_CABINET/X_all_you_need.pdf" } ], "images": [ { "url": "https://lh7-us.googleusercontent.com/sitesv-images-rt/APXyLQdOD7NG9EkcYJCVODj8ztJx_4CZYlL9BlSWGfQrLyg7Q4_VuYjQ24cp4PVAhIYNMWfLLoTz4M-WaK1zG3apGxaTQJv2lg7vRsB7hj1ZPnVzHj7njP3n8Q4RMtVWQPB0iaD8lO2E6oW-8y-UwjsjRdvZ9eADuR1ck7epbsGOhJ4-93eqwkUVu-5V1b1uEG1Bw2yYcJgqqgtTHh1q2SFgEHxit1BvfpnY1bs33aHV=w1280", "alt": "" } ], "embeds": [ { "path": "archive/embeds/spice-cabinet--all-you-need-1.html.txt", "text": "Matrix Rain with Blue Neon Text \"Those who ride the tiger can never dismount.\"", "links": [] } ], "iframes": [] } ], "assets": [ { "url": "https://lh3.googleusercontent.com/docsubipk/AP9E6xVvaconU2C_Ofxnz72Bewpok5L1VnUfe5tqmlYbt-JoeEtZzuttXTJEtastczMvTh2Q5fzxjmNl7xIYS9mQzybDLVUlih1hoiIOdLSjPCfGNfaul-_ChjdJsIexMzePbQveZuDb5A7sPac49PN6k6eJRcSLUitWsZzIfqGilm8vnguV8EldzaM", "local_path": "assets/original/7fb8bcec7d71d45b.jpg", "mime": "image/jpeg", "size": 29700, "sha256": "06f8e034a29b5b7a263b927e551084781c557b6bcc3b4ded9ea23b510719d9f2" }, { "url": "https://lh7-us.googleusercontent.com/sitesv-images-rt/AMxu72uYaoHUDmD8j3eM8MFUhwBCttZslmB8vlWn1atTFpFPyZ-pCoeiRizAcMYedgP4LopuIBW1fk2Vi4dS70y2tFf96lh0lY18wlbYAZK5hiPD4NeSYdxAMuAA6YT1jFxQ2ja6a8T7_FShNHGp8E3FFLCODEMHlJARDYZH9mo=w16383", "local_path": "assets/original/5144b4b5ce6b382b.jpg", "mime": "image/jpeg", "size": 504532, "sha256": "786bf25ead062990a353cd85cd565b3189ea8a45c5cd07109c4b6946123a8bbe" }, { "url": "https://lh7-us.googleusercontent.com/sitesv-images-rt/AMxu72ufa4Zp8Ru5pFkme5yxzeQtCvNhrkaW6HYlgizSCkLGbwSwtVw0AKQqM6iaCNvxLAOgoXzemzxlZ3EyP1EcxVS2CqHa4QrK5ZUm_uykUANYwyhMFKQFXRset89oO6AUCHb9deAj2hvqAL6iYZwH7O3L79mqG7TTG69nrmVBDM43vpWbuZNrIrG-qv40z1UV0t5FSBLFd7D7Vo5guyw=w1280", "local_path": "assets/original/765713cc658aa597.png", "mime": "image/png", "size": 515301, "sha256": "a5da7b2a323b41dbe7703cd713238a52316f48b70001e187c23ba7803db73df3" }, { "url": "https://lh7-us.googleusercontent.com/sitesv-images-rt/AMxu72uuOujrpAH1JTPaEKjZC138aKIkt96d3hHQZqRqgsEjk9QgXYEZ1ZqACdF1MJis4KWZN5sC-PJK9s5us1kqfRm2jlfIs_w5mJRKvwKez19X0l6Gy9Ib-vJfErLRPosSMCy4ifBOhI2TywI90UXFXniR-Q8MdRN4FkZSO-PCItab2Mb1k9VbA8oqrL_w6z6kXF4oI2DHEDVIK2Hf5fA=w1280", "local_path": "assets/original/c4b747c8813324f8.png", "mime": "image/png", "size": 1012936, "sha256": "3393b0d646d5ef9f91bc08ea8b4939a5ec0be55d72edae0eade0de155799ab8a" }, { "url": "https://lh7-us.googleusercontent.com/sitesv-images-rt/APXyLQcGg9_7yVENI-l6_G9kBlVfdQ7Nn7CcQVEzKhYjSLi6FpIMRZNaw75r8OMWGKGsM8eVu2jskcNwIsQFXPTGiFeuFIlRugHuTsO4ej88ZdOKEEWbgjJDLzGQWimnncvnd4mgLp3NnMF92znHCn6N5bc_AxZzN_MvKuUxqwUX9uxgZBpRxQgDIZTYzhqbds07T0qfS_Osb9nPVL2bxzPnkQPiaS0wtkU5f-FIw2GAHq0=w1280", "local_path": "assets/original/c68b6e4e81ec6494.png", "mime": "image/png", "size": 170350, "sha256": "d06730605cf41db655e16b8095e5729e6d80056c86cbfed7ef2c6e2f822a771a" }, { "url": "https://lh7-us.googleusercontent.com/sitesv-images-rt/APXyLQcLSE8ukg6PJgOf4qvWE5hA6Kk2E6QvyW7q5PaM4gOid_ETZVz-1s60D53HOTNQl-4CRNl0qWQ-wwgcEgAvQXCTS0nXLZM70mXe0CnFSM0BZXuvS_txWqaXmIbyL9rzliJjPz9c-7AZdFLP2y7OPc8MLcJ7DUhSfDhD2pyDZvL09DgqB4KApH2H2XpiKTc=w1280", "local_path": "assets/original/72822da0523e8439.png", "mime": "image/png", "size": 896040, "sha256": "c8c76140b72305c9b5437ff4cc5bf6c92922d8738e949e1ddb14aa29b2cc0368" }, { "url": "https://lh7-us.googleusercontent.com/sitesv-images-rt/APXyLQcVX0cAfYNyl0Wexx1uk7KwD43XMT2TyxJB9WTp0-oB5hBaN9u4DULrOwuPhDQUEVr3dwzXCaIcZtC3D_rzDXn9mtvRUM2yiQSSmYCuk0DgShFimBDOQUtGvG2Bbfu5RP37sEvH9tga3SxkDGknzzCdxbIStRP2cJ5IZ1aqKgqhYSRd0fFPFdCx9IOtf7rIj-mInBDI_BHIMF3MP4e9IbnhU6yY5DbKmtayDydKAR0=w1280", "local_path": "assets/original/a0c048dc15a50f68.png", "mime": "image/png", "size": 215078, "sha256": "38b0b3af879cdaa69e5b262420f1b6f6fbfb463c3b8c9c0f8867e9e6f4961488" }, { "url": "https://lh7-us.googleusercontent.com/sitesv-images-rt/APXyLQcVlyMsTJRQZigXDVUK8Qg1gDb0v49Uw2PirwuGhBGJNKeAyxzmsN5mApqcjUxz9oVKQ5SVze1hat6uvFAS8fzsrK5eoKqu4b4wB53X4peO7X6-vxNkr1PwKPsXneZasyUJZ3XzV1Ykel76eGZWE6iN0oPgS3Q2awhTef62Lswx-c2HaD_zWEZlAaVwtuynuYTs_TE-yzElgsVQpVs=w1280", "local_path": "assets/original/e562d4daf2ae48b4.jpg", "mime": "image/jpeg", "size": 430105, "sha256": "ef7efa90ba7aafa618871b095858bd8c321acb9c37a868b67fba9f198d844749" }, { "url": "https://lh7-us.googleusercontent.com/sitesv-images-rt/APXyLQcZ_awlFHvUUc_A6jRBi-zig3C94ebRnEKV4Vymz7oPrvcVrhapYUlhXNrNBna-WqHMUrOlogDks9MoFKGaUBWN_hNFrzMcKFlyWHNZVijbnZcrOobvP40IrnDx74moO1YwMTTchf6JM1xYmry9BTwf2QN7B2vrF_KpjccduQKyJMipgRMYnna6CQ9Ac7SoIcj3gOXfcQs26qDZUfnlcDUuxqts96EcBIP9jhfg=w1280", "local_path": "assets/original/39179dec09867791.png", "mime": "image/png", "size": 132446, "sha256": "bb856ea58cb90eafa61eefe010b939aab586b4080b2d80ad20bbea1f6c178069" }, { "url": "https://lh7-us.googleusercontent.com/sitesv-images-rt/APXyLQcbbreJ7LoKNs38zG-lfk6WJkhMV4km6t06-rqka60fR6gbplETLwkGGCk5QkuqU4E7BLMOhberAab0sHtDDFnz3WuIMDBS_NodE4R2rChMk20QOsWynf6Vjmkc9sBB6gCdnLB1vURbXXScVyc1GdeSqk5aHDbIagMUIB7Se8__C5jgNx3HrCTn0l5zwyc5J10VSfi1JSFhDaWowwc=w1280", "error": "HTTP Error 403: Forbidden" }, { "url": "https://lh7-us.googleusercontent.com/sitesv-images-rt/APXyLQceIaifsJwvEWis88ChFXgFXKk9OxMdliU--naJ_AAWk21MEuddEd7Yx41DtAlF7LUWDmd_lIY5Rh9fpf0yUzMHhrSgvc688oFKsKhFYENpf6zAT1Q69Zmveq-Fod1gk_PWa1rj7gWg3kE713BgkpJDsR706Wi6sEyWRtgTSagVc4LcPFwCH3nks06d56jeWQqyD93TfPLOWgSBLxkj_kuJUbqUWtRabSx6T--j=w1280", "local_path": "assets/original/fe1049584f77886b.jpg", "mime": "image/jpeg", "size": 29102, "sha256": "190e26fe419b9029287ab529da18ad97f2d7112a85ab9e1279eb83bc529daddf" }, { "url": "https://lh7-us.googleusercontent.com/sitesv-images-rt/APXyLQcgFfMgACYulXVN192_6cRcx0rBHbs9Y446C3eazjCPfP9I3s_q6ill1cWDa52c0R7COAiyeYBkjaJagRnSuDRqdV6MYySttf9YNZuA67Mvod4Vxh09hzcpVqNqXKZS2-TC8THOVImHlmhLe2kZCy5PJu2HwYJ3lC8ACd9BAVFUsmsvFclu3GZ14rotC98rrZyAhnRYFr0Xf7M-AEE=w1280", "local_path": "assets/original/080c73b9fbcbf3a2.jpg", "mime": "image/jpeg", "size": 20041, "sha256": "89e904eda42d70b03eba97603270b0b2176142ad7174087bb2a88e308e6438a2" }, { "url": "https://lh7-us.googleusercontent.com/sitesv-images-rt/APXyLQcrV8h_gCo3BZOOmGDVs52dEtcihGlXuraBDr-EYusXpChHjkvsT6CWdFVnc98m1Q1PfDctHfTOspEPIqDmhhCHivX6iyPaZLHd3cswlSVFWND2HZtGSoGylMBqpHREuES4hFIcpQwec3FxgL7k31-Sh6JKy6LTUjbp-4qelayVrENSIoJx8Wd6QMhNEhMWMYwgAKuw283W0ZvgTHg=w1280", "local_path": "assets/original/157aec3d79d1e200.png", "mime": "image/png", "size": 1028269, "sha256": "caf21004f895e9db777e62ee8190dcd2d250bf2c847043c4ba78eaa99ccc8c5a" }, { "url": "https://lh7-us.googleusercontent.com/sitesv-images-rt/APXyLQctNpVZ8XS1z27Qgl5wZnjbfzZLQdBHfzVPO3d0Pplj5Y4ubzZDWlN47oslQCpjKvqUbd4OfvHz57WVHXQdccE-MhREOBd9eiC5mab7X1ZYlptX_5Msdmb7nBN2PEC6cLV_Rr4DpgMuau9GryhvkaxGO_YQV0awzvLs-bFlQ1nv9X1DRQLWqqGCG0RcAe-3_DZ-O1oIJn2cjVkSYHs=w1280", "error": "HTTP Error 403: Forbidden" }, { "url": "https://lh7-us.googleusercontent.com/sitesv-images-rt/APXyLQcyT6qTA_xshDtcFDsKKXLO-nOmKeR4xDJ3nVc0UvjbN8ecSnM1M3kFbH-mZ5YmYDu3vlrv-FFvkUG8rUbTxOfjMHsQpB3tUg8-RyZo3n_TFqwnPN3lWjRIFDUSm6pwU462Xqe4rUtoY-OiI6M80dDOQzJu960stDpQzN9vMeVLdw2v7EO7QdkZRK1nJ8z_z0UkP9LhlbtoqYursLE=w1280", "local_path": "assets/original/2359ea25ed31569b.jpg", "mime": "image/jpeg", "size": 24733, "sha256": "6456e51ee57f13a1078149a37b10d8dfd39e9f8ee4c973c5b8b3711ea453d1b3" }, { "url": "https://lh7-us.googleusercontent.com/sitesv-images-rt/APXyLQd7LQGRXASuir3ZO8Bd-vxqckcbi1wyQnSToro6TyVKwz5N8pZpNMxOK6cUlDzC6zAL3SHrVGDYs8TOxye3-Q3O4cm0iMAMwp7vIpOYKqDe1pUC1P03UKNQ3MiZ8lzx04G6P2FFryfUu0Ev2uZllW5mMncIMFZ5jBOCnOgTjqMG5CGTp_ASdDriXqMuadrD9gYoIFgwbTx12hgNFOc=w1280", "error": "HTTP Error 403: Forbidden" }, { "url": "https://lh7-us.googleusercontent.com/sitesv-images-rt/APXyLQdOD7NG9EkcYJCVODj8ztJx_4CZYlL9BlSWGfQrLyg7Q4_VuYjQ24cp4PVAhIYNMWfLLoTz4M-WaK1zG3apGxaTQJv2lg7vRsB7hj1ZPnVzHj7njP3n8Q4RMtVWQPB0iaD8lO2E6oW-8y-UwjsjRdvZ9eADuR1ck7epbsGOhJ4-93eqwkUVu-5V1b1uEG1Bw2yYcJgqqgtTHh1q2SFgEHxit1BvfpnY1bs33aHV=w1280", "local_path": "assets/original/3f62a7dd451bbaed.png", "mime": "image/png", "size": 564357, "sha256": "2a44b3e0a31f5422a528a928d16a19be6e5a6c87106b270c75f98842d967dd88" }, { "url": "https://lh7-us.googleusercontent.com/sitesv-images-rt/APXyLQdTovYHTVJk6_0ohQZqFPtn623A248oQFOKINRdU1cAK5SM1xMSHTXrPLFOwroBqhTGO97pCVngWVCy9mta06OBPhsU3Q3TOSRe7uAbQwHc0F-t_pK85t5rzPkpaZPxJAuSygETCHKel3FPNdHnZwGqcqtPqn4p_MNH_2Gkv66O9bfq77EgOYD8dQuEww7PdI5wg2TiWCh3UF_PNb9KrZUtx7-y3qNeRTbgh8dUEPo=w1280", "error": "HTTP Error 403: Forbidden" }, { "url": "https://lh7-us.googleusercontent.com/sitesv-images-rt/APXyLQdeekk5lF7yRT-zmgs8fy360GM3lpOwy-j8VtG7J6DSIUd-nNTiGoX7NJATugcDe-u-bIZRtFbUPkSHFWpWichYweJhaKTME48VBKoc5BtVrZqYBKe_l0XLeBrP241fPj8fRuZZEz6N1XjsnRiaZHzrfGr4759gCqHCnlEFGFJ7XzmHu2xprJwbBfPyY_1q2oL3P8UbXUJYelY3zqjWpEhQR9ahx4kXm1MzCdz17uo=w1280", "local_path": "assets/original/8bf4aeeaf8b73e8e.jpg", "mime": "image/jpeg", "size": 102683, "sha256": "fc3bd2bde2929b0b60577dc21923c7936e6563a4c8fdf900088c8558f16d9d89" }, { "url": "https://lh7-us.googleusercontent.com/sitesv-images-rt/APXyLQdjSmRmC8--cLj7uZusAZgreCQUSZsejenxGXSStY1PPh4licBM-ed_xfbOJWD48jv3JD2uz2m7E6AGNEJMbwEnLXGH2LMRztdA6cjhGHkz8NK0zC8wL6VvR4K-Yt8OMi-O3fS6v4iGpJF8xohSx_vJNeSvO8L8EoocTUjZhPJy6bTI59djdV11locAma8p5BJHH4ve-Q517GmcLzkdHzxxOUqZQlNjWtcNgOGhkO4=w1280", "local_path": "assets/original/9aa8ac1f3b54e258.png", "mime": "image/png", "size": 105302, "sha256": "5ecc67153b5dd1647016a374d5e3fdbc9f9e25000050e90a3cb59661cbbf678b" }, { "url": "https://lh7-us.googleusercontent.com/sitesv-images-rt/APXyLQdmxLu6Gr7AdFbUl2XcvlfuiGg1L-Wk15BuP2R_NFJ5l5DHUcxPTH0BynRbMg1efN0RW8IxfzNsumVu9Iyradc8a83SlezTV93pI3WaUMJFJ6nLvmswqwt4yee8TqSweKe5Su0zYuyEdPJF0YyEE4ZpW670UxFm7YpoPCk_at2Ba4xiBNU4PRvraiA-frqYJIK0I9foKSaj04bZLMsbeQv7cSohBk9zVTi_Om_7PRc=w1280", "error": "HTTP Error 403: Forbidden" }, { "url": "https://lh7-us.googleusercontent.com/sitesv-images-rt/APXyLQdy9Zw4XJCUgBuJ366sG_IhMoZ6KTqBukfS4sa0NT83iVAlrPD8gOcnoUZLew1U30BlhAs4qAW6EHgGYni-THMz9rfhrXDKSyC_fhanm2bUzklZa7xcFihlngmBO46u79ePGTBorLwSRJ-bd96lKbq73yV0QDB6wQUzII4l7h9bxUc3S5LZrSGk_nAkT_YSC5lkq9NJkKCBB7Lmy6Q4ocHcaBwV0o82dMbwIJUjcBM=w1280", "local_path": "assets/original/c0982773633cb852.png", "mime": "image/png", "size": 381420, "sha256": "7a904dbd163c1d6b472ccd8f79c0e962a5bf4508b41790787eb68d25b7ffb2c3" }, { "url": "https://lh7-us.googleusercontent.com/sitesv-images-rt/APXyLQeDElCONCOk1p4pjB-9mlQV63Gj9NFPNwUifo1sQt9KIiIGuOrzFKTz1VLq2-npJxcQrb3x5C1MQrZbrWnGf2qsMdYG437HzSY3Bl7REGCM7njFD_KUUVKbP2t9ncWIyZGFkLFClL3k-Fq7P2eO3AFvT315WEBpHLKbi5vtriXRXBVkDaB83EGER4QgEKoIISQJX32X3GbT9WQKBcVVH_IRulyjVdy8dArvi5KbxRM=w1280", "error": "HTTP Error 403: Forbidden" }, { "url": "https://lh7-us.googleusercontent.com/sitesv-images-rt/APXyLQeDfVbyxmCKSkJTw5mDTlIOBZAkDTfYRiAsagmeiGmPC0pqG89HaLg2bN8u6xNPAjIoq2Wp1M4ttxO9ynzNR0Y1mCOQTTQ6iUnhTS8IYZTIHid2B69qVQILwFImbKrWEZYs0nD0lxU0H7RgUJhCwREfrXaeacUft8RStYE=w16383", "error": "HTTP Error 403: Forbidden" }, { "url": "https://lh7-us.googleusercontent.com/sitesv-images-rt/APXyLQeONqa5xm1_qdeVulp0eMUR1qFhTCQyDwsiZOPiVjGiea0BBhP6RFi7uq3tk2sWwdZiksweJNXjcz0e_rMG1PSLWGI5_MCEMu5jls8HdOpoNM8RW3vO4z6_pbSAwWuidt9ROi2vEdDUyKa2Hy67o33Jr-q_CmrSfuQNYljQCtwhPm7lVA0nxXNdhc4fVRWjkT28mt04D7wWqw7LCobtum6eOAT7oihTuIw6jFJNijs=w1280", "error": "HTTP Error 403: Forbidden" }, { "url": "https://lh7-us.googleusercontent.com/sitesv-images-rt/APXyLQeZUHK1IGqHVYPrZf2eAtu15LQoWGBwRN2Pws7XqvjvZH2Lnop4q1P9mVXYJnVz0G_OSX-3JkMyIdBgiQ3OxIhDhJV2smxAMjUqysU8ocrr2n9JeUgXuZrwCVy463OilzNs7wxyRH0zNyGg9joN31QZN49hQ_IBkKw9o61GSjtmor8436MnFheKOg7kFb5OENcaxJFNd43sob-Xke8=w1280", "error": "HTTP Error 403: Forbidden" }, { "url": "https://lh7-us.googleusercontent.com/sitesv-images-rt/APXyLQeh4BwJSU_VFHKsbzP7ZLxZ7W5trWZ3lamfd-duLF6zrUor0OeavybW3MeZQxNvTSaAALA0VHSqwpETcgcgHvD5gtbioH3XsHv7QrzEb4QktlMpym_dYasOBF7KiHynsWEom0uc52Cs5DvFjCFlu6zru2P5w2gPgQIEUU0=w16383", "local_path": "assets/original/d7909bf86dea2c8b.jpg", "mime": "image/jpeg", "size": 504532, "sha256": "786bf25ead062990a353cd85cd565b3189ea8a45c5cd07109c4b6946123a8bbe" }, { "url": "https://lh7-us.googleusercontent.com/sitesv-images-rt/APXyLQekqEoLP-t7RMLb2_a9bGJOq439Fj7-Yq9aymOuc-N6PmT6_t2F-TujqDJRvqwlqSJ-3Fu7l2cPXiNdEv0F6FVzjrUTsc25aO695FM5ftXFWy_WNMbD1YKs827t1rLCN3JnDVGf9KhWFsx2TlRM5mmvsM9o5AvaWLoAAb0CiJ_UlL6XVwVOS0GNPf32JYjyQ_7zEUqWrfc9Ss5LKqPainY_Evfpt1I8tSYnOk6l=w1280", "local_path": "assets/original/9f0ae239b6a4a02c.png", "mime": "image/png", "size": 547439, "sha256": "eeced638f6ef436f3e6f406f4dd5ed668ad635ab14c893d4f40af6402d734b32" }, { "url": "https://lh7-us.googleusercontent.com/sitesv-images-rt/APXyLQf2omiISf8EOPNVmOQdRFOHBeNdoAN_NoN1qwS_PGoe_ZEi3cXrDmJk40rLOOVuZkOVaSElreR0hmXtgtDcyhPT4lxsDtDXHn6KFPJ3OR1dfdaXXynSL7kb1pJwrOwFjIajIqa5OS2ms18-J9CpskqPPg9HKg6jjjJsS8mZ70hHRtrB5Lt6U2mJFUDjLAy1GlZe0jI4egNmloYKWKEZ9Ds7CAV_7uvgFQWiXBKswK8=w1280", "local_path": "assets/original/e61a9fe099f55459.jpg", "mime": "image/jpeg", "size": 224564, "sha256": "7479fb201e9512bba856042e7f3bdf7a5f2fb35fcac4a18040d8042b0c7bdc90" }, { "url": "https://lh7-us.googleusercontent.com/sitesv-images-rt/APXyLQf8PiIUD3foTh4VVP_cpVmM48Y7NpV4-0tM_vehtfZm_1THEBWoVgflhBIH47siXqGwmTze8imheRYnPPRT09xkQ2iHUDkCtbsxlk6DK0_ApLxZN29y_PzOoMIC2MnESF93adUCIaT0ZceWn21EoygwYsrOrW9J6aFdMeuUlU3_nCALLPaPk2iX0OZxM3R09BfUUwvZBUL0tmh0=w1280", "error": "HTTP Error 403: Forbidden" }, { "url": "https://lh7-us.googleusercontent.com/sitesv-images-rt/APXyLQf8n7Hfem6UjywE4TqsPeSK_ul4teTkQKD6KdMuUYlHdCJyhAVCBAN2qw_tyi2LBH24g4Mdhh8HUd_lDrp4kVJ1-gmchmCTx99WeeQUYv6mgdpfx72VVpO0K5nOq1baCzh-GU9Uvf4R_oLozPMy_1peybc1Xv9Uq9hbKCRReIVg733Erv-D_gaaU3CTCCRWOi0_RbS_ONSpVFbUp3rHpGmDzRfCg-DKACscw63mXvs=w1280", "local_path": "assets/original/437cc4b4de474ae9.jpg", "mime": "image/jpeg", "size": 15193, "sha256": "1d5afb56eb7ecbf9a819211ba99bbbe6060694951497f61bbd85c2a25c839ccf" }, { "url": "https://lh7-us.googleusercontent.com/sitesv-images-rt/APXyLQfSjW8dMZ94qFQvjO0Oy0FGH6VmSJa0XSchiWIpLWchwb7VXBUtzrxzW_WmqHmGOIWZcE5eqa0h-qF1NdAgOmRDxy6lTPRWCwCciDcqeEpZgWtI31Wkts2pUCOveQMYC5f0HGL1oDtl_7aqMZjLFTcjg792PhBdQU4QjjayiL0_UaREAgIUV94YYURPsrFRmJX-o5v6N-KrmMZclgSgjpYvszulTqok1GI4PV92EXU=w1280", "error": "HTTP Error 403: Forbidden" }, { "url": "https://lh7-us.googleusercontent.com/sitesv-images-rt/APXyLQffzJQaZyUgrriVznNgccmQx6geL4kIJSqPiBJIVYMlAx7xg8S1dCF1pftsvOmxKq6V7m5aLqs5Pb0XRtbqUhqDl9s8qBAKLt9JBMhaVZWJt0xK305xxT679KVrtb5RZsMi-EqjWNWWPJ5lhfB6CdWHgat8vfjfmHRZisDApDV_vysHuwt-Tf1aw0Osg84R9w7IXYyWpbCjaoJdjxI=w1280", "local_path": "assets/original/8dd62ef26b972e74.jpg", "mime": "image/jpeg", "size": 143071, "sha256": "394f36dad47ae2330ccc7249e3f4d7516f86932b130b0fc2bc91710c70b6aea7" }, { "url": "https://lh7-us.googleusercontent.com/sitesv-images-rt/APXyLQfnUiZYlNvH13Lf_DhDV2NF86-kP_1SPX4wHuIlY2hMlchRPA7mo4cII8BRrSo_J8QsWWIMEcbyZjowGn6V19kHNUZy4vxeLRQmCms9MDS_B7BqpBltAYB1bAuUzqyslK0Om5a0IiKESnXchSyHg4aSlYuuyxfBgZfNX4SRkiMkjoXkH1VjlJRwsYMizwLOygEJn_asbRxmIqSI1NxRxkte0ZuVKCbu8X_x4nUcJcU=w1280", "error": "HTTP Error 403: Forbidden" } ], "failures": [], "scope": "All publicly linked same-domain pages and their directly embedded images. External links and iframe URLs retained; third-party websites and video streams are not mirrored." }