Papers of the day   All papers

It’s Hard For Neural Networks to Learn the Game of Life

Comments

hardmaru: It's Hard for Neural Networks To Learn the Game of Life They investigate how neural nets learn the rules of Game of Life and why they often miss finding the correct solution. Highlights limitations of gradient-based learning https://arxiv.org/abs/2009.01398 blog: https://bdtechtalks.com/2020/09/16/deep-learning-game-of-life/ https://t.co/k6z9xJIQCK

7 replies, 742 likes


Gary Marcus: It's Hard for Neural Networks To Learn the Game of Life; new paper that examines how some “neural networks rely on lucky random initial weights of subnetworks called “lottery tickets” that converge quickly to a solution”. https://arxiv.org/abs/2009.01398

1 replies, 96 likes


Turil Cronburg: Every different pattern finds different patterns in other patterns. We tend to forget this. One way to think about it is to realize that each system — animal, vegetable, mineral, etc. — is basically a unique mathematical function, so each outcome will almost always be unique.

2 replies, 8 likes


mark: https://arxiv.org/abs/2009.01398

2 replies, 7 likes


arxiv: It's Hard for Neural Networks To Learn the Game of Life. http://arxiv.org/abs/2009.01398 https://t.co/h1MiYTRmaO

0 replies, 6 likes


Dmytro Mishkin: It’s Hard For Neural Networks to Learn the Game of Life Jacob M. Springer, Garrett T. Kenyon https://arxiv.org/abs/2009.01398.pdf tl;dr: Initialization and overparameterization is everything for the task in question. https://t.co/rJwrLsR6wb

0 replies, 6 likes


Toby Walsh (Hiring 4 PostDocs + 8 PhDs): The title of this paper should be thought of literally and metaphorically! "It's Hard for Neural Networks To Learn the Game of Life" https://arxiv.org/abs/2009.01398

0 replies, 1 likes


Content

Found on Oct 01 2020 at https://arxiv.org/pdf/2009.01398.pdf

PDF content of a computer science paper: It’s Hard For Neural Networks to Learn the Game of Life