Developing Neural Algorithmic Reasoning

By Petar Veličković et al
Published on Dec. 10, 2022
Read the original document by opening this link in a new tab.

Table of Contents

Motivation
What do we mean by algorithm ?
Why algorithms?
The core problem
Neuralising an algorithm
Breaking the bottleneck
Neural Algorithmic Reasoning
Why do we need a new field?
What do we know, theoretically?
What do we know, empirically?
The CLRS-30 Benchmark
Representation
Representation: encoding
Representation: processing
Representation: decoding

Summary

This document explores the development of Neural Algorithmic Reasoning, discussing the concept of algorithms, the importance of algorithms, and the challenges in algorithmic reasoning. It delves into neuralising algorithms to tackle core problems and breaking bottlenecks. The document also introduces the concept of Neural Algorithmic Reasoning and the need for a new field in this area. Theoretical and empirical knowledge, along with benchmarks like CLRS-30, are presented, highlighting the representation, processing, and decoding aspects of algorithmic reasoning.
×
This is where the content will go.