What Do Latent Action Models Actually Learn?
This is an incremental analysis for researchers in unsupervised video representation learning, focusing on theoretical insights rather than practical improvements.
The paper investigates whether latent action models (LAMs) learn action-relevant changes or noise from unlabeled videos, using a tractable linear model to analyze connections to PCA and strategies like data augmentation.
Latent action models (LAMs) aim to learn action-relevant changes from unlabeled videos by compressing changes between frames as latents. However, differences between video frames can be caused by controllable changes as well as exogenous noise, leading to an important concern -- do latents capture the changes caused by actions or irrelevant noise? This paper studies this issue analytically, presenting a linear model that encapsulates the essence of LAM learning, while being tractable.This provides several insights, including connections between LAM and principal component analysis (PCA), desiderata of the data-generating policy, and justification of strategies to encourage learning controllable changes using data augmentation, data cleaning, and auxiliary action-prediction. We also provide illustrative results based on numerical simulation, shedding light on the specific structure of observations, actions, and noise in data that influence LAM learning.