Papers of the day   All papers

Strong Generalization and Efficiency in Neural Programs

Comments

Yujia Li: Excited to share our new work on learning to discover neural programs that strongly generalize and outperform hand-coded algorithms for sorting (quick sort) and other tasks, in number of execution steps. https://arxiv.org/abs/2007.03629 w/ @FelixAxelGimeno @pushmeet @OriolVinyalsML https://t.co/S0N3xWVxRE

13 replies, 737 likes


hardmaru: Their learned programs can outperform hand-coded programs in terms of efficiency on several algorithmic tasks, such as sorting, searching in ordered lists and a version of the 0/1 knapsack problem, while also generalizing to instances of arbitrary length. https://arxiv.org/abs/2007.03629

2 replies, 219 likes


John Regehr: TIL people are doing program synthesis research but they are calling it "neural program induction" (and getting cool results too) https://arxiv.org/pdf/2007.03629.pdf

1 replies, 61 likes


Zach Vorhies: Computer Science: This AI just coded a sorting solution that was faster than quick sort. https://arxiv.org/abs/2007.03629

5 replies, 35 likes


Hacker News: Neural programmer better than Quicksort https://arxiv.org/abs/2007.03629

0 replies, 24 likes


Pushmeet Kohli: Our latest work on algorithm induction from our team @DeepMind. With @liyuajia, @FelixAxelGimeno and @OriolVinyalsML

0 replies, 12 likes


elvis: Could this be useful to prepare for technical interviews?

0 replies, 8 likes


Saad Khan: AI is going to replace basic/mundane (non-scientific) coding. It’s not just gpt3 but lots of recent NLP papers on code generation (e.g. https://arxiv.org/abs/2007.03629) ... 4/6

1 replies, 6 likes


Hector Yee: Getting closer to "Fire upon the deep" by Vernor Vinge where people persuade computers to do things instead of program them

1 replies, 5 likes


Kévin Dunglas: “Our learned model can perform sorting perfectly on any input data size we tested on, with O(nlogn) complexity, whilst outperforming hand-coded algorithms, including quick sort, in number of operations.” https://arxiv.org/abs/2007.03629

0 replies, 5 likes


pheras: "our learned model can perform sorting perfectly on any input data size we tested on, with O(nlogn) complexity, whilst outperforming hand-coded algorithms, including quick sort, in number of operations even for list sizes far beyond those seen during training"

1 replies, 5 likes


Hacker News 250: Neural programmer better than Quicksort https://arxiv.org/abs/2007.03629 (http://news.ycombinator.com/item?id=23788298)

0 replies, 3 likes


Ward Plunet: Machine learning for the win.

0 replies, 3 likes


Angsuman Chakraborty: Neural programmer better than Quicksort https://arxiv.org/abs/2007.03629

0 replies, 1 likes


Saurabh Bhatnagar: Cool development. But also makes me sad as a developer.

0 replies, 1 likes


Content

Found on Jul 09 2020 at https://arxiv.org/pdf/2007.03629.pdf

PDF content of a computer science paper: Strong Generalization and Efficiency in Neural Programs