Published Sep 10, 2021

Episode 13: Jonathan Frankle, MIT, on the lottery ticket hypothesis and the science of deep learning

Explore deep learning's evolving landscape with MIT's Jonathan Frankle as he delves into the Lottery Ticket Hypothesis, critiques current AI methodologies, and advocates for a scientific reevaluation in neural network research.

Episode Highlights

Topics covered

Deep Learning Challenges
01
Scientific Methodology
02
Neural Network Pruning
03

Episode 13: Jonathan Frankle, MIT, on the lottery ticket hypothesis and the science of deep learning

Topics covered

Popular Clips

AI Impact Evolution

Network Optimization Insights

Academic Challenges

Neural Network Experiments

Embracing Machine Learning Messiness

Network Pruning Insights

Neural Network Discoveries

Pruning Techniques Revisited

Researcher's Software Engineering Secrets

Rethinking Review Criteria

Learning from Mistakes

Future of Deep Learning

GPU Infrastructure and Experiment Management

Training Phases

Navigating Technology Policy

Episode Highlights

Deep Learning Challenges

Scientific Methodology

Neural Network Pruning

Episode 13: Jonathan Frankle, MIT, on the lottery ticket hypothesis and the science of deep learning

Topics covered

Popular Clips

Episode Highlights

Deep Learning Challenges

Scientific Methodology

Neural Network PruningJonathan Frankle discusses the Lottery Ticket Hypothesis, exploring its origins and implications for neural network training. He critiques existing pruning techniques, advocating for more rigorous scientific practices in machine learning research.

Neural Network Pruning