Dropout-based Active Learning for Regression

TL;DR

This paper introduces a fast active learning method for regression using dropout-based uncertainty estimation in neural networks, enabling efficient data sampling to improve models without high computational costs.

Contribution

It presents a novel, computationally efficient active learning algorithm for regression that leverages dropout uncertainty, suitable for large-scale neural network applications.

Findings

Comparable or better performance than baselines

Effective on synthetic and real-world datasets

Generalizable to other deep learning architectures

Abstract

Active learning is relevant and challenging for high-dimensional regression models when the annotation of the samples is expensive. Yet most of the existing sampling methods cannot be applied to large-scale problems, consuming too much time for data processing. In this paper, we propose a fast active learning algorithm for regression, tailored for neural network models. It is based on uncertainty estimation from stochastic dropout output of the network. Experiments on both synthetic and real-world datasets show comparable or better performance (depending on the accuracy metric) as compared to the baselines. This approach can be generalized to other deep learning architectures. It can be used to systematically improve a machine-learning model as it offers a computationally efficient way of sampling additional data.