Human-Guided Complexity-Controlled Abstractions
Not on the map. Why?
Abstract (arXiv)
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.
Key facts. Cited by 1 paper in the map (data through 7 Oct 2026). Builds on 21 of its 38 references that are in the map. First posted on arXiv 26 Oct 2023.
Builds on
21 of 38 references.
References · 21
- 2015Deep Residual Learning for Image Recognitioncited by 51,809
- 2017Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networkscited by 4,861
- 2017Neural Discrete Representation Learningcited by 4,650
- 2017Fashion-MNIST: a Novel Image Dataset for Benchmarking Machine Learning Algorithmscited by 4,283
- 2016Categorical Reparameterization with Gumbel-Softmaxcited by 2,708
- 2016The Concrete Distribution: A Continuous Relaxation of Discrete Random Variablescited by 1,135
- 2016Deep Variational Information Bottleneckcited by 915
- 2015Deep Learning and the Information Bottleneck Principlecited by 793
- 2018This Looks Like That: Deep Learning for Interpretable Image Recognitioncited by 731
- 2019Meta-Learning with Implicit Gradientscited by 523
- 2017Opening the Black Box of Deep Neural Networks via Informationcited by 494
- 2018Learning to Adapt in Dynamic, Real-World Environments Through Meta-Reinforcement Learningcited by 270
9 more
- 2017Deep Learning for Case-Based Reasoning through Prototypes: A Neural Network that Explains Its Predictionscited by 234
- 2021Deformable ProtoPNet: An Interpretable Image Classifier Using Deformable Prototypescited by 120
- 2019Interpretable and Steerable Sequence Learning via Prototypescited by 72
- 2018Theory and Experiments on Vector Quantized Autoencoderscited by 55
- 2022ProtoVAE: A Trustworthy Self-Explainable Prototypical Variational Modelcited by 34
- 2022Representation Learning in Deep RL via Discrete Information Bottleneckcited by 13
- 2018Variational Information Bottleneck on Vector Quantized Autoencoderscited by 5
- 2021Visual communication of object concepts at different levels of abstractioncited by 5
- 2022Prototype Based Classification from Hierarchy to Fairnesscited by 4
Built on by
Citing papers per year: 2024: 1.
1 paper.