SADL method reduces representation collapse in AI models by matching supply and demand
SADL: On the Demand-Supply Geometry of Representation Collapse
RoboticsMachine Learning
Summary
AI models that learn by comparing inputs sometimes end up representing everything too similarly, which is called representation collapse. The authors found that this happens because the part of the model that encodes the inputs shares too much common info across data points. They propose a new method called SADL that keeps track of how much difference the later parts of the model expect and how much the encoder can deliver, adding a penalty only when needed. This leads to more stable training and better performance in tasks like control and image classification.
What this means in practice
- •For machine learning engineers: Build more stable control system models by preventing representation collapse during training with the SADL method.
- •For computer vision developers: Improve accuracy and training stability of image classification models by applying SADL to manage encoder representation differences.
Authors
Junhan Sun, Yigui Feng, Mingzhe Liu, Haoyan Xu, Lei Ye, Ziqi Han, Hanru Bai, Yanzhao Zhang, Hao Zhao, Guofeng Zhang
Abstract
In Joint-Embedding Predictive Architectures (JEPAs), a well-constrained projector output does not by itself prevent representation collapse. We find that, from initialization, the encoder representation is dominated by a component shared by different inputs. A batch-normalized projector removes this component, so regularization after the projector cannot correct it, and the differences between representations shrink until they collapse. We propose SADL (Supply And Demand Loss), which keeps the original training objective and constrains the encoder directly when needed. SADL compares the representation differences that the projector output demands with the differences the encoder can retain and pass on to the projector, and adds a constraint on the encoder only when supply falls short of demand. It works with different Demand Regularizers and introduces no auxiliary networks, additional trainable parameters, or architectural changes. In single-task and jointly trained four-task LeWorldModel (LeWM) and INTACT experiments, SADL lowers the collapse rate over ten random seeds from 50% to 0%. After one epoch of training, encoder effective rank is 3.4-6.4 times that of the corresponding baseline without SADL, and latent prediction error falls to 1/20-1/303 of it. Without additional tuning, SADL stays robust as its loss weight spans orders of magnitude, and with several Demand Regularizers INTACT exceeds 95% mean search-free control success. On four image datasets, SADL also consistently improves accuracy and stability across random seeds. These results show that matching the representation differences a regularizer demands with the differences the encoder can supply leads to faster and more stable representation learning in control and image classification.