Human-Guided Complexity-Controlled Abstractions
This addresses the issue of poor generalization in neural networks for rapid model finetuning, leveraging human insight, but it is incremental as it builds on existing abstraction and complexity control methods.
The paper tackles the problem of neural networks learning task-specific latent representations that fail to generalize, by training models to generate a spectrum of discrete representations with controlled complexity. The results show that tuning representation complexity to a task-appropriate level supports the highest finetuning performance with only a small number of labeled examples, and users can identify this level using visualizations.
Neural networks often learn task-specific latent representations that fail to generalize to novel settings or tasks. Conversely, humans learn discrete representations (i.e., concepts or words) at a variety of abstraction levels (e.g., "bird" vs. "sparrow") and deploy the appropriate abstraction based on task. Inspired by this, we train neural models to generate a spectrum of discrete representations, and control the complexity of the representations (roughly, how many bits are allocated for encoding inputs) by tuning the entropy of the distribution over representations. In finetuning experiments, using only a small number of labeled examples for a new task, we show that (1) tuning the representation to a task-appropriate complexity level supports the highest finetuning performance, and (2) in a human-participant study, users were able to identify the appropriate complexity level for a downstream task using visualizations of discrete representations. Our results indicate a promising direction for rapid model finetuning by leveraging human insight.