On the Stepwise Nature of Self-Supervised Learning
This provides insight into the fundamental mechanisms of self-supervised learning, which is incremental in understanding training dynamics for researchers in machine learning.
The paper tackles the training process of joint embedding self-supervised learning methods, finding that they learn high-dimensional embeddings one dimension at a time in discrete steps, as demonstrated through a linearized model and experiments with deep ResNets using Barlow Twins, SimCLR, and VICReg losses.
We present a simple picture of the training process of joint embedding self-supervised learning methods. We find that these methods learn their high-dimensional embeddings one dimension at a time in a sequence of discrete, well-separated steps. We arrive at this conclusion via the study of a linearized model of Barlow Twins applicable to the case in which the trained network is infinitely wide. We solve the training dynamics of this model from small initialization, finding that the model learns the top eigenmodes of a certain contrastive kernel in a stepwise fashion, and obtain a closed-form expression for the final learned representations. Remarkably, we then see the same stepwise learning phenomenon when training deep ResNets using the Barlow Twins, SimCLR, and VICReg losses. Our theory suggests that, just as kernel regression can be thought of as a model of supervised learning, kernel PCA may serve as a useful model of self-supervised learning.