Petar Veličković (DeepMind) — Neural Algorithmic Reasoning

Jakub Zavrel (Founder and CEO at Zeta Alpha) interviews Petar Veličković (Staff Research Scientist at DeepMind) about their recent paper on the CLRS Algorithmic Reasoning Benchmark.

📄 Paper "The CLRS Algorithmic Reasoning Benchmark" https://arxiv.org/abs/2205.15659
❓ About us: https://www.zeta-alpha.com/

Abstract: Learning representations of algorithms is an emerging area of machine learning, seeking to bridge concepts from neural networks with classical algorithms. Several important works have investigated whether neural networks can effectively reason like algorithms, typically by learning to execute them. The common trend in the area, however, is to generate targeted kinds of algorithmic data to evaluate specific hypotheses, making results hard to transfer across publications, and increasing the barrier of entry. To consolidate progress and work towards unified evaluation, we propose the CLRS Algorithmic Reasoning Benchmark, covering classical algorithms from the Introduction to Algorithms textbook. Our benchmark spans a variety of algorithmic reasoning procedures, including sorting, searching, dynamic programming, graph algorithms, string algorithms and geometric algorithms. We perform extensive experiments to demonstrate how several popular algorithmic reasoning baselines perform on these tasks, and consequently, highlight links to several open challenges. Our library is readily available at https://github.com/deepmind/clrs

Timestamps:
00:00 Introduction, getting started in AI
06:20 What is Neural Algorithmic Reasoning?
10:42 The CLRS Algorithmic Reasoning Benchmark paper
15:40 How do you encode an algorithmic problem into a benchmark?
19:10 What does it mean to achieve a certain score in this benchmark? F1 score
21:15 Validation vs. Test set score discrepancy
23:00 Connections with Transfer Learning
24:40 Public Leaderboards on this benchmark
27:40 Is Deep Learning inherently compatible or incompatible with algorithmic reasoning?
31:30 OpenAI's Codex, AlphaCode and the difference between program induction and program synthesis
34:30 Where do Quantum Computers fit in this picture
36:25 How do you manage to keep track of the literature?
38:51 Outro