MISSION: Ultra Large-Scale Feature Selection using Count-Sketches
This addresses the problem of interpretable feature selection for machine learning practitioners dealing with massive datasets, representing a novel method rather than an incremental improvement.
The paper tackles the challenge of feature selection on ultra large-scale datasets with billions of dimensions, where standard methods fail due to memory constraints, and presents MISSION, a framework that uses Count-Sketch to achieve efficient feature selection with O(log^2(p)) memory while retaining interpretability.
Feature selection is an important challenge in machine learning. It plays a crucial role in the explainability of machine-driven decisions that are rapidly permeating throughout modern society. Unfortunately, the explosion in the size and dimensionality of real-world datasets poses a severe challenge to standard feature selection algorithms. Today, it is not uncommon for datasets to have billions of dimensions. At such scale, even storing the feature vector is impossible, causing most existing feature selection methods to fail. Workarounds like feature hashing, a standard approach to large-scale machine learning, helps with the computational feasibility, but at the cost of losing the interpretability of features. In this paper, we present MISSION, a novel framework for ultra large-scale feature selection that performs stochastic gradient descent while maintaining an efficient representation of the features in memory using a Count-Sketch data structure. MISSION retains the simplicity of feature hashing without sacrificing the interpretability of the features while using only O(log^2(p)) working memory. We demonstrate that MISSION accurately and efficiently performs feature selection on real-world, large-scale datasets with billions of dimensions.