Learnable: Theory vs Applications

TL;DR

This paper compares applied machine learning and PAC learning theory, showing how PAC theory can inform practical problems but often requires impractically large training sets, and advocates clarifying misconceptions to better align theory with practice.

Contribution

It formalizes the connection between applied learning and PAC theory, highlighting limitations and proposing to clarify misconceptions for better practical relevance.

Findings

PAC theory can guide applied learning under certain conditions

Large training sets are required by PAC theory, limiting practicality

Shedding misconceptions can improve alignment between theory and practice

Abstract

Two different views on machine learning problem: Applied learning (machine learning with business applications) and Agnostic PAC learning are formalized and compared here. I show that, under some conditions, the theory of PAC Learnable provides a way to solve the Applied learning problem. However, the theory requires to have the training sets so large, that it would make the learning practically useless. I suggest shedding some theoretical misconceptions about learning to make the theory more aligned with the needs and experience of practitioners.