Vector Institute
Vector Institute
  • 334
  • 148 327
Hassan Ashtiani - Differentially Private Learning of Distributions
Vector Faculty Affiliate Hassan Ashtiani talks about some recent developments in differentially private distribution learning. These include advances in basic problems of hypothesis selection, learning Gaussians, and learning Gaussian Mixture Models. Through these examples Hassan shows some of the shared challenges in designing private estimators, as well as some generic frameworks (such as private-to-non-private reductions) for addressing them.
Speaker's Bio: Hassan Ashtiani is an Assistant Professor in the Department of Computing and Software at McMaster University and a faculty affiliate at Vector Institute, Toronto. He obtained his Ph.D. in Computer Science in 2018 from University of Waterloo where he was advised by Shai Ben-David. Broadly speaking, a major theme in his research is the design and analysis of sample-efficient learning methods that are robust to (i) privacy-related attacks, (ii) model misspecification, (iii) adversarial attacks, and/or (iv) distribution shift.
Переглядів: 58

Відео

Purang Abolmaesumi - Machine Learning Technologies for Accessible Point of Care Ultrasound
Переглядів 2314 днів тому
Talk abstract: Point-of-care ultrasound (POCUS) describes the process whereby any healthcare practitioner uses a portable ultrasound device to diagnose conditions at the bedside, whether in an urban hospital, rural clinic, or a remote location. When used appropriately, POCUS can reduce wait times for diagnostic tests, make patient treatment pathways more efficient, streamline transfers to save ...
Nicolas Loizou - Next-GenAdaptive Optimization Algorithms for Large-Scale Machine Learning Models
Переглядів 3314 днів тому
Abstract: Stochastic gradient descent (SGD) is the workhorse for training modern large-scale supervised machine learning models. In this talk, we will discuss recent developments in the convergence analysis of SGD and propose efficient and practical adaptive variants for faster convergence. We will start by presenting a novel adaptive (no tuning needed) learning rate for SGD. We will introduce ...
Agustinus Kristiadi - Uncertainty-Guided Optimization on LLM Search Tree
Переглядів 7514 днів тому
Abstract: Beam search is a standard tree search algorithm when it comes to finding sequences of maximum likelihood, for example, in the decoding processes of large language models. However, it is myopic since it does not take the whole path from the root to a leaf into account. Moreover, it is agnostic to prior knowledge available about the process: For example, it does not consider that the ob...
Navigating Careers in AI for Black & Indigenous Students, with Black & Indigenous Leaders
Переглядів 5421 день тому
This engaging and informative webinar aims to inspire and educate Black & Indigenous students about the vast career opportunities in AI and ML. Our panel featured accomplished Black & Indigenous professionals who have carved out successful careers in these fields. They shared their personal journeys, discussed the roles they play in their respective organizations, and offered insights into how ...
Digital Transformation - FastLane 2.0 - Vector Intern Talks
Переглядів 2821 день тому
Manoj and Surabie's talk explores how automation platforms can be leveraged to automate manual, repetitive tasks and improve the current workflows of the FastLane program. For the project, custom reports on Salesforce, dashboards on PowerBI and automated workflows on Zapier were built to efficiently track onboarding and event engagement data. The goal of the automation project was to increase e...
VLMs for Medical Report Generation and VQA Using CT Scans - Vector Intern Talks
Переглядів 9421 день тому
Mohammed’s talk explores vision-language models for medical report generation and visual question answering using CT scans. Specifically, in the talk, Mohammed details his journey of participating in the "MICCAI24 AMOS-MM: Abdominal Multimodal Analysis Challenge," exploring methodological and post-processing techniques to the improve performance and outlining current issues with the tasks. Adva...
CanDoo: Reshaping Legal Research With AI - Vector Intern Talks
Переглядів 2421 день тому
Jack and Sammed's presentation describes Candoo, a legal research tool using AI to help lawyers and researchers find better information, faster. A new feature of CanDoo, Query Composer, is introduced to aid users enhance their database queries. In the past, creating boolean queries could be challenging due to their confusing syntax and rigid nature. But with Query Composer, users now have the p...
Revolutionizing Investment Analysis: The Next-GenAI Summarization Tool - Vector Intern Talks
Переглядів 5021 день тому
The presentation examines the demanding and time-consuming task of summarizing quarterly reports from investment managers. The strategy implemented includes fine-tuning, prompt engineering, and retrieval-augmented generation (RAG) using agentic RAG and LangChain. Tools such as LlamaParse for text extraction and OpenAI’s LLM GPT-4o for summarization were utilized. The primary objective is to cre...
FL4Health: Driving Adoption of a Federated Learning Platform - Vector Intern Talks
Переглядів 5321 день тому
Janice's talk explores the challenges and opportunities of implementing federated learning (FL) in healthcare. She delves into FL4Health, an open-source library designed to streamline FL experimentation and deployment in healthcare settings. The presentation highlights strategies for overcoming adoption barriers, including product usability and securing buy-in from senior management. Janice dis...
Reload - Selective Data Forgetting and Replacement - Vector Intern Talks
Переглядів 1521 день тому
Machine Unlearning is an important task in machine learning surrounding making a model selectively ‘forget’ part of its training data. In some cases, the data may be unavailable, and conventional methods require access to unlearn. Those that do not, do not perform well. In this talk I introduce Reload, a new method I developed with Dr. Rahul Krishnan which is inspired by memorization in neural ...
Partner Portal Search & Recommendation Enhancements - Vector Intern Talks
Переглядів 3821 день тому
The goal of this project was to deliver more relevant content (and search results) to users across different industries and technical expertise levels on Vector's Partner Portal website. To achieve this, I deployed a Flask API on GCP that received search queries and user details from the portal, and returned tailored content recommendations and more insight search results. The API leveraged a Q...
Vector Use Case Catalog: A RAG Knowledge Base Implementation - Vector Intern Talks
Переглядів 7321 день тому
Eric Edem Dzeha and Nikki Smith's project focuses on the development and implementation of a Retrieval-Augmented Generation (RAG) solution tailored for Vector Institute’s diverse range of AI projects. This system efficiently catalogs and provides easy access to use cases from past projects, ensuring that relevant information can be retrieved on demand. The project involved structuring textual d...
Vector for Canada: The Future of AI Leadership
Переглядів 2053 місяці тому
Canada is a global AI leader, boasting a unique combination of talent, investment, and entrepreneurial spirit. Outperforming many of our G7 peers in AI talent, research publications, and venture capital, we've made significant strides. However, we face a crucial choice: Should we push the boundaries, invest more, and integrate AI across our economy to remain at the forefront? Or should we act c...
Vector Faculty Affiliate Talk - April Khadeni
Переглядів 653 місяці тому
April Khademi is a Faculty Affiliate at the Vector Institute and Canada Research Chair in AI for Medical Imaging. She is an Associate Professor of Biomedical Engineering at Toronto Metropolitan University and Principal Investigator of the Image Analysis in Medicine Lab (IAMLAB). Her research, funded by various Canadian institutions, focuses on AI algorithms for medical imaging. She is also an A...
Vector Faculty Affiliate Talk - Nisarg Shah
Переглядів 864 місяці тому
Vector Faculty Affiliate Talk - Nisarg Shah
Vector Lecture Series - Anima Anandkumar
Переглядів 4194 місяці тому
Vector Lecture Series - Anima Anandkumar
Fine Tuning and Alignment for Large Language Models - Vector Intern Talks
Переглядів 1944 місяці тому
Fine Tuning and Alignment for Large Language Models - Vector Intern Talks
Prompt Engineering Lab - Vector Intern Talks
Переглядів 1444 місяці тому
Prompt Engineering Lab - Vector Intern Talks
Identifying Attack Campaigns from Network Alerts - Vector Intern Talks
Переглядів 514 місяці тому
Identifying Attack Campaigns from Network Alerts - Vector Intern Talks
Odyssey: A Toolkit to Build Foundation Models Using Electronic Health Records - Vector Intern Talks
Переглядів 754 місяці тому
Odyssey: A Toolkit to Build Foundation Models Using Electronic Health Records - Vector Intern Talks
Digital Booklet: Anomaly Detection - Vector Intern Talks
Переглядів 1014 місяці тому
Digital Booklet: Anomaly Detection - Vector Intern Talks
Public Health and Fairness Evaluation using ML Models on Clinical Notes - Vector Intern Talks
Переглядів 554 місяці тому
Public Health and Fairness Evaluation using ML Models on Clinical Notes - Vector Intern Talks
Partial Weight Exchange in Federated Learning - Vector Intern Talks
Переглядів 784 місяці тому
Partial Weight Exchange in Federated Learning - Vector Intern Talks
Interpretability and Bias Identification in LMs - Vector Intern Talks
Переглядів 504 місяці тому
Interpretability and Bias Identification in LMs - Vector Intern Talks
Vector Lecture Series - Elliot Paquette
Переглядів 2175 місяців тому
Vector Lecture Series - Elliot Paquette
Vector Visitor Talk - Neil Lawrence
Переглядів 4575 місяців тому
Vector Visitor Talk - Neil Lawrence
The Vector Institute - AI can be Remarkable
Переглядів 5695 місяців тому
The Vector Institute - AI can be Remarkable
Vector Lecture Series - Renjie Liao
Переглядів 1646 місяців тому
Vector Lecture Series - Renjie Liao
Planning Your Path: How To Become a Leader in Data & AI - Full Webinar
Переглядів 1656 місяців тому
Planning Your Path: How To Become a Leader in Data & AI - Full Webinar

КОМЕНТАРІ

  • @jeelanshahtlyr6076
    @jeelanshahtlyr6076 22 дні тому

    MASHA ALLAH EXCELLENT WORK KEEP UP THE GOOD WORK BROTHER

  • @CamilleButler-u2o
    @CamilleButler-u2o 22 дні тому

    Taylor Dorothy Hall Eric Lewis Edward

  • @MShahbazKharal
    @MShahbazKharal 29 днів тому

    I searched "Prafulla Dhariwal" on UA-cam to find his interviews/talks and I found this gem. GPT4o's amazing voice mode is definitely the realization of Prafulla's slide at ua-cam.com/video/Jlb1IQ9ooxw/v-deo.html.

  • @Renallyson289
    @Renallyson289 3 місяці тому

    😂

  • @AryaRishi
    @AryaRishi 4 місяці тому

    Hi from India, I am really concerned will AI do the same to us what we did to animals ?

  • @markteague8889
    @markteague8889 4 місяці тому

    Ok, from the get go ... Dr. Hinton's assertion that hardware and software are fundamentally different is an incorrect one. While this may be true in practice; in theory, the difference between hardware and software is only a logical one. What is implemented in software today can be implemented in hardware tomorrow. This boundary between hardware and software is a fluid one with new microarchitecture designs removing complex functions that are deemed unnecessary and implementing new ones.

  • @chinthakuntlabalu9026
    @chinthakuntlabalu9026 4 місяці тому

    Thankyou universe ❤❤❤888

  • @masternobin
    @masternobin 4 місяці тому

    Here after open AI launched chat gpt+4o

  • @JustincoShaw
    @JustincoShaw 5 місяців тому

    thank you

  • @bhaskartripathi
    @bhaskartripathi 5 місяців тому

    Very interesting use case. We did a similar thing for Joe Fresh at Loblaws

  • @LPB2016
    @LPB2016 7 місяців тому

    😅

  • @DivineMisterAdVentures
    @DivineMisterAdVentures 8 місяців тому

    The Vector Institute has according to my screen in the U.S., 2.87k subscribers, 97 views, and I'm the first commenter, in 60 days. If you are in Canada, what data do you have?

  • @001HK0
    @001HK0 8 місяців тому

    It's always interesting to see model theory pop up in stat decision theory - as an observer of the former (from afar) it has been interesting to see the work being done on tameness as well.

  • @LPB2016
    @LPB2016 8 місяців тому

    Hi I think your work is so good

    • @LPB2016
      @LPB2016 8 місяців тому

      Your work is so good I wish I could do

  • @johanneskiessling4085
    @johanneskiessling4085 9 місяців тому

    Hello Michael & Caroline. To me this seems to be very important work which you are doing. Best regards from Germany.

  • @LPB2016
    @LPB2016 9 місяців тому

    Hi😊❤

  • @LPB2016
    @LPB2016 9 місяців тому

    Hi ❤

  • @afsheenmaroof6209
    @afsheenmaroof6209 9 місяців тому

    can i get the code detail for prompt engineering please

  • @ivan8960
    @ivan8960 10 місяців тому

    Horrible, contrived idea and drastically increases chances of a bad AI future.

  • @hanankhimish5502
    @hanankhimish5502 10 місяців тому

    Informative presentation.

  • @gcarbo
    @gcarbo 11 місяців тому

    Excellent content, but unfortunately, Dina is so painful to listen to. And it is not her voice timbre or accent.

  • @SamratDe-yq5go
    @SamratDe-yq5go 11 місяців тому

    Hello Vector Institute Owner, I saw your UA-cam channel, you upload very nice videos on your channel. I have watched all the videos and noticed that all your videos are beautiful. But, there are many problems with your channel. Your main problem is that the video views are not coming in and your videos are not reaching the target audience. Because your channel is not SEO! If you don't fix this problem, your video will not get video views organically. Because UA-cam's algorithm works on keywords, but your channel has no keywords! This problem can only be solved by SEO. If your channel SEO is done then your videos will rank on UA-cam and help your videos reach the right people. This will increase your video views. Remember, SEO is not just one job, it is organized into many activities. If you want to do SEO I can help you. I am a UA-cam SEO expert, I can solve your problem. Thank you Have a nice day

  • @Manuel-nk1bm
    @Manuel-nk1bm 11 місяців тому

    😒 *Promo SM*

  • @nast697
    @nast697 11 місяців тому

    Nice talk, Shohreh.

  • @macknightxu2199
    @macknightxu2199 Рік тому

    hi how to choose from HE,DP,SMPC for Federated learning for privacy protection? Any standards to follow?

  • @a.c.e7407
    @a.c.e7407 Рік тому

    Great conference, just wish the audio was better.

  • @xaviermachiavelli5236
    @xaviermachiavelli5236 Рік тому

    2

  • @azalkhan1045
    @azalkhan1045 Рік тому

    Thanks, great work.

  • @Rahul-xo8di
    @Rahul-xo8di Рік тому

    Hi I am starting my freshman year. So can u guide me about, how should I approach to have a high paying career in AI. PS :- I do know anything even about coding. Just about to start

  • @famcivilengineer34
    @famcivilengineer34 Рік тому

    Truly remarkable Innovation

  • @periklisspanos7185
    @periklisspanos7185 Рік тому

    With those things can we win the mega millions or not If is not why we needed them hypotheses are bad guesses

  • @MichaelM28
    @MichaelM28 Рік тому

    He's a man and always will be

  • @moreisdifferent
    @moreisdifferent Рік тому

    I think this series of videos don't get the attention they deserve, and understandably so. I wish Vector had multiple UA-cam channels for the different kinds of contents, such as project presentations, socials, keynote-style talks, career advice (like this video) etc. What I'm seeing is infrequent publication of several videos at once with unpredictable focus areas, which deters viewers. The thumbnail isn't very descriptive either. I highly appreciate all the content Vector publishes, but I think it can be published with better planning.

  • @souvikmajumdar3183
    @souvikmajumdar3183 Рік тому

    Excellent helpful for my research area... It will be more fruitful we can get the code details a little bit...

  • @AzinAsgarian-kf4cn
    @AzinAsgarian-kf4cn Рік тому

    Great summary and work, thanks Saeed!

  • @Mahm00d27
    @Mahm00d27 Рік тому

    Well done @Taiwo

  • @sundayamos9174
    @sundayamos9174 Рік тому

    Very good presentation

  • @marekdudek1714
    @marekdudek1714 Рік тому

    @10:00 sound like "cortical columns" or "1000 brains theory" (by Numenta) .. now we are getting somewhere :)

  • @rb8049
    @rb8049 Рік тому

    Wow! Hinton described Trumps tweets just as I understood how they work. My understanding of language is synchronization of brains. There is very little information in spoken or written communication. What is happening is the brains are synchronized in context and understanding. Only a few people seem to understand in even a partial sense what the brain does. Glad he is leading us there in understanding.

  • @joshuasansom_software
    @joshuasansom_software Рік тому

    Thanks for the summary. Helped me understand things little better.

  • @SubhamKumar-eg1pw
    @SubhamKumar-eg1pw Рік тому

    The github link doesn't exists

    • @vectorinstituteai
      @vectorinstituteai Рік тому

      Links from the video: github.com/google-research/prompt-tuning github.com/orgs/VectorInstitute/repositories Thank you.

  • @xue8364
    @xue8364 Рік тому

    Was this the fall event?

  • @tessuz
    @tessuz Рік тому

    Thanks man

  • @selinyucelbak
    @selinyucelbak Рік тому

    thank you so much for this, I have learned a lot from this lecture

  • @ryanharvey485
    @ryanharvey485 Рік тому

    ρгό𝔪σŞm 😎

  • @richardj.haldane3936
    @richardj.haldane3936 2 роки тому

    Always a step in tomorrow....Vevtor

  • @carlacrafford180
    @carlacrafford180 2 роки тому

    FASCINATING!!! At present I am working with the opposite - all analogue, all to do with interpretations of perspective (as seen) and mirrors (some warped on purpose). I would love to make contact if possible. Exhibition coming up...

  • @ranrana3
    @ranrana3 2 роки тому

    Could you share the reference for this talk?

  • @alphatech_ai
    @alphatech_ai 2 роки тому

    Well done, Ade Adeoye

  • @Bokaseca
    @Bokaseca 2 роки тому

    Thanks, now I understand