[2006.10782] AI Feynman 2.0: Pareto-optimal symbolic regression exploiting graph modularityopen searchopen navigation menucontact arXivsubscribe to arXiv mailings

We present an improved method for symbolic regression that seeks to fit data to formulas that are Pareto-optimal, in the sense of having the best accuracy for a given complexity. It improves on the previous state-of-the-art by typically being orders of magnitude more robust toward noise and bad data, and also by discovering many formulas that stumped previous methods. We develop a method for discovering generalized symmetries (arbitrary modularity in the computational graph of a formula) from gradient properties of a neural network fit. We use normalizing flows to generalize our symbolic regression method to probability distributions from which we only have samples, and employ statistical hypothesis testing to accelerate robust brute-force search.

10 mentions: @tegmark@jaguring1@yoshi_and_aki@tjmlab
Date: 2020/06/26 17:21

Referring Tweets

@jaguring1 おっ「AIファインマン2.0」が登場! データから、それを満たす法則を自動発見するAI。1.0はファインマン物理学に載ってる100の法則を全て発見し、より難しい問題設定でも最高性能を15%から90%にしたが、2.0ではノイズに強くなり、記号的密度推定を含むより困難な問題も解けた t.co/ZhFGrsDb51
@yoshi_and_aki 今日の積ん読論文消化。AIファインマン2.0。与えられた複雑度のモデルの中で最高精度になるパレート最適なモデルを選んでくるらしい。でもモデルの候補としてニューラルネットから記号にもってくる探索はブルートフォースだし、局所最適になりそうな気がする、、、t.co/gNruJbrAJW
@tegmark Got data? The code described in our new paper auto-discovers simple and accurate formulas describing it by exploiting modules and generalized symmetry. Let me know if you'd like to collaborate on applications! force search. t.co/0vhIWtowgz t.co/1X4fssi96U

Related Entries

Read more [1711.08028] Recurrent Relational Networks
0 users, 4 mentions 2018/12/05 22:45
Read more [1909.11942] ALBERT: A Lite BERT for Self-supervised Learning of Language Representations
0 users, 7 mentions 2019/09/27 23:18
Read more [2003.01200] Natural Language Processing Advancements By Deep Learning: A Surveycontact arXivarXiv T...
0 users, 5 mentions 2020/03/04 02:20
Read more [1907.00205v4] The Ramanujan Machine: Automatically Generated Conjectures on Fundamental Constantsop...
0 users, 3 mentions 2020/05/27 02:21