Quang-Huy Nguyễn
Quang-Huy Nguyễn
  • 16
  • 1 315
Open Ended and AI Generating Algorithms in the Era of Foundation Models - Jeff Clune
[Vanderbilt Machine Learning] Open Ended and AI Generating Algorithms in the Era of Foundation Models - Jeff Clune (University of British Columbia), December 09, 2024
Abstract: Foundation models (e.g. large language models) create exciting new opportunities in our longstanding quests to produce open-ended and Al-generating algorithms, wherein agents can truly keep innovating and learning forever. In this talk I will share some of our recent work harnessing the power of foundation models to make progress in these areas. I will cover our recent work on OMNI (Open-endedness via Models of human Notions of Interestingness), Video Pre-Training (VPT), Thought Cloning, Automatically Designing Agentic Systems, and The Al Scientist.
Переглядів: 121

Відео

Concept Learning Across Domains and Modalities - Jiajun Wu, Stanford University1:01:31Concept Learning Across Domains and Modalities - Jiajun Wu, Stanford University
Concept Learning Across Domains and Modalities - Jiajun Wu, Stanford University
Переглядів 242Місяць тому
[Vanderbilt Machine Learning] Concept Learning Across Domains and Modalities - Jiajun Wu (Stanford University), November 04, 2024 Abstract: I will discuss a concept-centric paradigm for building agents that can learn continually and reason flexibly across multiple domains and input modalities. The concept-centric agent utilizes a vocabulary of neuro-symbolic concepts. These concepts, including ...
Automatic Data Curation for Self-Supervised Learning: A Clustering-Based Approach - Huy Vo (Meta AI)1:04:39Automatic Data Curation for Self-Supervised Learning: A Clustering-Based Approach - Huy Vo (Meta AI)
Automatic Data Curation for Self-Supervised Learning: A Clustering-Based Approach - Huy Vo (Meta AI)
Переглядів 272 місяці тому
[Vanderbilt Machine Learning Seminar Series] Automatic Data Curation for Self-Supervised Learning: A Clustering-Based Approach - Huy Vo (Meta AI) - October 28th Abstract: Self-supervised features are the cornerstone of modern machine learning systems. They are typically pre-trained on data collections whose construction and curation typically require extensive human effort. This manual process ...
Memory Mosaics - Leon Bottou1:06:20Memory Mosaics - Leon Bottou
Memory Mosaics - Leon Bottou
Переглядів 522 місяці тому
[Vanderbilt Machine Learning Seminar Series] Memory Mosaics - Leon Bottou (Meta AI (FAIR) in New York) - August 26th, 2024 Abstract: Memory Mosaics are networks of associative memories working in concert to achieve a prediction task of interest. Like transformers, memory mosaics possess compositional capabilities and in-context learning capabilities. Unlike transformers, memory mosaics achieve ...
Distributional Preference Alignment of Large Language Models via Optimal Transport - Youssef Mroueh1:00:08Distributional Preference Alignment of Large Language Models via Optimal Transport - Youssef Mroueh
Distributional Preference Alignment of Large Language Models via Optimal Transport - Youssef Mroueh
Переглядів 352 місяці тому
[Vanderbilt Machine Learning Seminar Series] Distributional Preference Alignment of Large Language Models via Optimal Transport - Youssef Mroueh (IBM Research/MIT-IBM Watson AI Lab) - October 21, 2024 Abstract: Current LLM alignment techniques use pairwise human preferences at a sample level, and as such, they do not imply an alignment on the distributional level. We propose in this paper Align...
Generalizing Outside the Training Distribution through Compositional Generation - Yilun Du1:07:48Generalizing Outside the Training Distribution through Compositional Generation - Yilun Du
Generalizing Outside the Training Distribution through Compositional Generation - Yilun Du
Переглядів 1152 місяці тому
[Vanderbilt Machine Learning Seminar Series] Generalizing Outside the Training Distribution through Compositional Generation - Yilun Du (Google DeepMind / Harvard University) - September 23, 2024 Abstract: Generative AI has led to stunning successes in recent years but is fundamentally limited by the amount of data available. This is especially limiting in the embodied setting - where an agent ...
Unsupervised Learning Segmentation Of, By, and For Visual Recognition - Stella Yu1:07:29Unsupervised Learning Segmentation Of, By, and For Visual Recognition - Stella Yu
Unsupervised Learning Segmentation Of, By, and For Visual Recognition - Stella Yu
Переглядів 237 місяців тому
[Vanderbilt Machine Learning Seminar Series] Unsupervised Learning Segmentation Of, By, and For Visual Recognition - Stella Yu (University of Michigan) - November 14th, 2023 Abstract: Image segmentation in computer vision has evolved such that it is routinely treated as an end task. For example, for autonomous driving, we are interested in segmenting a road scene into (cars, bikes, motorcycles,...
The Art of Writing 1st Rank Conference in Computer Science - Prof. My Thai (University of Florida)1:31:41The Art of Writing 1st Rank Conference in Computer Science - Prof. My Thai (University of Florida)
The Art of Writing 1st Rank Conference in Computer Science - Prof. My Thai (University of Florida)
Переглядів 287 місяців тому
The Art of Writing 1st Rank Conference in Computer Science - Prof. My Thai (University of Florida)
Denoising as a Building Block for Imaging, Inverse Problems, and Machine Learning - Peyman Milanfar1:01:04Denoising as a Building Block for Imaging, Inverse Problems, and Machine Learning - Peyman Milanfar
Denoising as a Building Block for Imaging, Inverse Problems, and Machine Learning - Peyman Milanfar
Переглядів 557 місяців тому
[Vanderbilt Machine Learning Seminar Series] Denoising as a Building Block for Imaging, Inverse Problems, and Machine Learning - Peyman Milanfar (Google Research) - February 5th, 2024 Denoising is one of the oldest problems in imaging. There are thousands of papers on this topic, and their scope is vast and the approaches so diverse that putting them in some order (as I will do) is both useful ...
Diffusion Models for Scientific Discovery - Stefano Ermon1:03:08Diffusion Models for Scientific Discovery - Stefano Ermon
Diffusion Models for Scientific Discovery - Stefano Ermon
Переглядів 1197 місяців тому
Diffusion models are at the core of many state-of-the-art generative AI systems for media content such as images, videos, and audio. Due to their excellent sample quality and theoretical guarantees, they are emerging as an important tool in many scientific, medical, and engineering applications. In this talk I will present several extensions of diffusion models tailored to the unique challenges...
Characterizing Machine Unlearning through Definitions and Implementations - Nicolas Papernot1:02:47Characterizing Machine Unlearning through Definitions and Implementations - Nicolas Papernot
Characterizing Machine Unlearning through Definitions and Implementations - Nicolas Papernot
Переглядів 1377 місяців тому
[Vanderbilt Machine Learning Seminar Series] Characterizing Machine Unlearning through Definitions and Implementations - Nicolas Papernot (University of Torondo/Vector Institude) - March 18th, 2024 Abstract: The talk presents open problems in the study of machine unlearning. The need for machine unlearning, i.e., obtaining a model one would get without training on a subset of data, arises from ...
Understanding The Computational Bases of Robust Object Recognition In Humans and DNN - Frank Tong59:17Understanding The Computational Bases of Robust Object Recognition In Humans and DNN - Frank Tong
Understanding The Computational Bases of Robust Object Recognition In Humans and DNN - Frank Tong
Переглядів 87 місяців тому
[Vanderbilt Machine Learning Seminar Series] Understanding The Computational Bases of Robust Object Recognition In Humans and Deep Neural Networks - Frank Tong (Vanderbilt University) - March 25th, 2024 Abstract: Deep neural networks (DNNs) trained on object classification provide the best current models of human vision, with accompanying claims that they have attained or even surpassed human-l...
KAN: Kolmogorov Arnold Networks - Ziming Liu1:34:56KAN: Kolmogorov Arnold Networks - Ziming Liu
KAN: Kolmogorov Arnold Networks - Ziming Liu
Переглядів 2567 місяців тому
Abstract: Inspired by the Kolmogorov-Arnold representation theorem, we propose Kolmogorov-Arnold Networks (KANs) as promising alternatives to Multi-Layer Perceptrons (MLPs). While MLPs have fixed activation functions on nodes ("neurons"), KANs have learnable activation functions on edges ("weights"). KANs have no linear weights at all every weight parameter is replaced by a univariate function ...
Conformal prediction under ambiguous ground truth - David Stutz57:12Conformal prediction under ambiguous ground truth - David Stutz
Conformal prediction under ambiguous ground truth - David Stutz
Переглядів 38Рік тому
[Vanderbilt Machine Learning Seminar Series] Conformal prediction under ambiguous ground truth - David Stutz (Google DeepMind) - November 06th, 2023 Abstract: In safety-critical classification tasks, conformal prediction allows to perform rigorous uncertainty quantification by providing confidence sets including the true class with a user-specified probability. This generally assumes the availa...
How to Detect Out-of-Distribution Data in the Wild? - Sharon Y. Li1:14:01How to Detect Out-of-Distribution Data in the Wild? - Sharon Y. Li
How to Detect Out-of-Distribution Data in the Wild? - Sharon Y. Li
Переглядів 47Рік тому
[Vanderbilt Machine Learning Seminar Series] How to Detect Out-of-Distribution Data in the Wild? Challenges, Research Progress, and Path Forward - Sharon Y. Li. October 10th, 2023 Presenter: Sharon Y. Li [Assistant Prof. at Univeristy of Wisconsin-Madison]
From Seeing to Doing: Understanding and Interacting with the Real World - Fei-Fei Li1:05:24From Seeing to Doing: Understanding and Interacting with the Real World - Fei-Fei Li
From Seeing to Doing: Understanding and Interacting with the Real World - Fei-Fei Li
Переглядів 19Рік тому
From Seeing to Doing: Understanding and Interacting with the Real World - Fei-Fei Li