# Zero-bias autoencoders and the benefits of co-adapting features

arXiv: https://arxiv.org/abs/1402.3337 · HTML: https://mlsanity.com/paper/1402.3337/

## Key facts

- Cited by 15 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 13 Feb 2014.

## Built on by (most cited first, 15 of 15)

- [Scaling and evaluating sparse autoencoders](https://mlsanity.com/paper/2406.04093/) (2024)
- [Jumping Ahead: Improving Reconstruction Fidelity with JumpReLU Sparse Autoencoders](https://mlsanity.com/paper/2407.14435/) (2024)
- [Efficient Dictionary Learning with Switch Sparse Autoencoders](https://mlsanity.com/paper/2410.08201/) (2024)
- [Route Sparse Autoencoder to Interpret Large Language Models](https://mlsanity.com/paper/2503.08200/) (2025)
- [Residual Stream Analysis with Multi-Layer SAEs](https://mlsanity.com/paper/2409.04185/) (2024)
- [Decoding Dark Matter: Specialized Sparse Autoencoders for Interpreting Rare Concepts in Foundation Models](https://mlsanity.com/paper/2411.00743/) (2024)
- [The Persian Rug: solving toy models of superposition using large-scale symmetries](https://mlsanity.com/paper/2410.12101/) (2024)
- [PolySAE: Modeling Feature Interactions in Sparse Autoencoders via Polynomial Decoding](https://mlsanity.com/paper/2602.01322/) (2026)
- [Piecewise Linear Units Improve Deep Neural Networks](https://mlsanity.com/paper/2108.00700/) (2021)
- [Steered Generation via Gradient Descent on Sparse Features](https://mlsanity.com/paper/2502.18644/) (2025)
- [Sparsity Hurts: Simple Linear Adapter Can Boost Generalized Category Discovery](https://mlsanity.com/paper/2605.08183/) (2026)
- [Domain Adaptation for Facial Expression Classifier via Domain Discrimination and Gradient Reversal](https://mlsanity.com/paper/2106.01467/) (2021)
- [Bayesian optimization for sparse neural networks with trainable activation functions](https://mlsanity.com/paper/2304.04455/) (2023)
- [UnIT: Scalable Unstructured Inference-Time Pruning for MAC-efficient Neural Inference on MCUs](https://mlsanity.com/paper/2507.07885/) (2025)
- [Steered Generation via Gradient-Based Optimization on Sparse Query Features](https://mlsanity.com/paper/2605.23040/) (2026)

## Abstract (arXiv)

Regularized training of an autoencoder typically results in hidden unit biases that take on large negative values. We show that negative biases are a natural result of using a hidden layer whose responsibility is to both represent the input data and act as a selection mechanism that ensures sparsity of the representation. We then show that negative biases impede the learning of data distributions whose intrinsic dimensionality is high. We also propose a new activation function that decouples the two roles of the hidden layer and that allows us to learn representations on data with very high intrinsic dimensionality, where standard autoencoders typically fail. Since the decoupled activation function acts like an implicit regularizer, the model can be trained by minimizing the reconstruction error of training data, without requiring any additional regularization.