Reasoning-Modulated Representations
arXiv:2107.08881
Abstract
Neural networks leverage robust internal representations in order to generalise. Learning them is difficult, and often requires a large training set that covers the data distribution densely. We study a common setting where our task is not purely opaque. Indeed, very often we may have access to information about the underlying system (e.g. that observations must obey certain laws of physics) that any "tabula rasa" neural network would need to re-learn from scratch, penalising performance. We incorporate this information into a pre-trained reasoning module, and investigate its role in shaping the discovered representations in diverse self-supervised learning settings from pixels. Our approach paves the way for a new class of representation learning, grounded in algorithmic priors.
To appear at LoG 2022. 17 pages, 5 figures
References in corpus (14)
- Neural Message Passing for Quantum Chemistry
- Language Models are Few-Shot Learners
- Reconciling modern machine learning practice and the bias-variance trade-off
- A simple neural network module for relational reasoning
- Learning to Simulate Complex Physics with Graph Networks
- Agent57: Outperforming the Atari Human Benchmark
- Object-Centric Learning with Slot Attention
- Relational Neural Expectation Maximization: Unsupervised Discovery of Objects and their Interactions
- Contrastive Learning of Structured World Models
- Spatial Broadcast Decoder: A Simple Architecture for Learning Disentangled Representations in VAEs
- Pointer Graph Networks
- XLVIN: eXecuted Latent Value Iteration Nets
- Learning Object-Centric Video Models by Contrasting Sets
- Persistent Message Passing