Analyzing Optimization and Generalization in Deep Learning via Dynamics of Gradient Descent
Explore deep learning optimization and generalization through gradient descent dynamics, focusing on implicit preconditioning and its impact on matrix completion problems.
Deep Robust Reinforcement Learning and Regularization
Explore robust reinforcement learning, addressing uncertainties in deep RL through regularization techniques. Gain insights into robust MDPs, policy evaluation, and uncertainty-aware Bellman equations.
The AI Economist- Improving Equality and Productivity with AI-Driven Tax Policies
Explore AI-driven tax policies for improved economic equality and productivity, featuring deep learning, reinforcement learning, and multi-agent simulations in a simplified economic model.
Backpropagation and Deep Learning in the Brain
Explore deep learning in the brain, backpropagation, and credit assignment algorithms. Discover neuroscience evidence and new models of neurons in this computational brain theory talk.
The Information Bottleneck Theory of Deep Neural Networks
Explore the Information Bottleneck Theory's application to deep neural networks, examining statistical learning, information theory, and stochastic gradient descent in brain data analysis.
Explaining Landscape Connectivity of Low-Cost Solutions for Multilayer Nets
Explore landscape connectivity in multilayer networks, focusing on low-cost solutions, equivalent local minima, overparameterization, and dropout stability. Gain insights into non-convex optimization and deep neural networks.
From Classical Statistics to Modern Machine Learning
Explore the evolution from classical statistics to modern machine learning, examining key concepts like generalization, interpolation, and the "double descent" phenomenon in supervised learning.
Generative Models
Explore generative models in deep learning, covering theoretical perspectives, hacker models, information flow, and provable algorithms for learning classifiers and phylogenetic reconstruction.
Implicit Regularization I
Explore implicit regularization in deep learning, covering boosting, complexity control, optimization landscapes, and stochastic gradient descent for effective model training.
Adversarial Examples
Explore the intriguing world of adversarial examples in deep learning, their implications, and theoretical frameworks with expert Sébastien Bubeck from Microsoft Research.
Recent Developments in Over-parametrized Neural Networks, Part I
Explore recent developments in over-parametrized neural networks, focusing on optimization challenges, theoretical problems, and strategies for achieving optimal results in deep learning.
Generalization II
Explore advanced concepts in deep learning, including random averages, losses, and network architectures, with experts from UC Berkeley and MIT.
On the Foundations of Deep Learning - SGD, Overparametrization, and Generalization
Explore foundations of deep learning, focusing on SGD, overparametrization, and generalization. Delve into optimization dynamics, margin theory, and the impact of network architecture on performance.
Learning and Generalization in Over-parametrized Neural Networks - Going Beyond Kernels
Explore neural network learning and generalization beyond kernel methods, examining tangent kernels, learning algorithms, and complexity in over-parametrized networks.
Nearest Neighbors I
Explore nearest neighbor algorithms in machine learning, covering regression, classification, consistency, and convergence rates in metric spaces. Gain insights into statistical learning frameworks and smoothness conditions.
Generalization I
Explore deep learning generalization with experts Peter Bartlett and Sasha Rakhlin, covering perceptron algorithms and key concepts in machine learning theory.
First-Order Stochastic Optimization
Explore first-order stochastic optimization techniques for machine learning, covering gradient descent, adaptive learning, and convergence rates in data science applications.
Flexible Neural Networks and the Frontiers of Meta-Learning
Explore cutting-edge approaches in meta-learning and flexible neural networks for rapid skill acquisition in AI agents, with applications in few-shot image classification and real-world scenarios.
Studying Generalization in Deep Learning via PAC-Bayes
Explore PAC-Bayes theory's application to deep learning generalization, focusing on risk bounds, optimal priors, and SGD prediction for neural networks.
Using Deep Learning for Perception in Autonomous Systems - A Perspective from Control Theory
Explore deep learning for perception in autonomous systems, focusing on efficient navigation using monocular RGB cameras in unknown environments. Control theory insights and practical lessons included.
Lessons Learned from Evaluating the Robustness of Defenses to Adversarial Examples
Explore robustness of AI defenses against adversarial examples, evaluating threat models and defense strategies to improve security in deep learning systems.