Perception Learning: A Formal Separation of Sensory Representation Learning from Decision Learning
This work addresses the problem of decoupling perception from decision-making in machine learning, offering a foundational approach with potential broad impact, though it is incremental in formalizing existing ideas.
The paper introduces Perception Learning (PeL), a paradigm that separates sensory representation learning from decision learning by optimizing task-agnostic signals for properties like stability and informativeness, and proves that PeL updates are orthogonal to task-risk gradients.
We introduce Perception Learning (PeL), a paradigm that optimizes an agent's sensory interface $f_φ:\mathcal{X}\to\mathcal{Z}$ using task-agnostic signals, decoupled from downstream decision learning $g_θ:\mathcal{Z}\to\mathcal{Y}$. PeL directly targets label-free perceptual properties, such as stability to nuisances, informativeness without collapse, and controlled geometry, assessed via objective representation-invariant metrics. We formalize the separation of perception and decision, define perceptual properties independent of objectives or reparameterizations, and prove that PeL updates preserving sufficient invariants are orthogonal to Bayes task-risk gradients. Additionally, we provide a suite of task-agnostic evaluation metrics to certify perceptual quality.