A Novel Active Learning Regression Framework for Balancing the Exploration-Exploitation Trade-Off.

Dina Elreedy,Samir I Shaheen,Amir F Atiya

doi:10.3390/e21070651

Abstract

Recently, active learning is considered a promising approach for data acquisition due to the significant cost of the data labeling process in many real world applications, such as natural language processing and image processing. Most active learning methods are merely designed to enhance the learning model accuracy. However, the model accuracy may not be the primary goal and there could be other domain-specific objectives to be optimized. In this work, we develop a novel active learning framework that aims to solve a general class of optimization problems. The proposed framework mainly targets the optimization problems exposed to the exploration-exploitation trade-off. The active learning framework is comprehensive, it includes exploration-based, exploitation-based and balancing strategies that seek to achieve the balance between exploration and exploitation. The paper mainly considers regression tasks, as they are under-researched in the active learning field compared to classification tasks. Furthermore, in this work, we investigate the different active querying approaches—pool-based and the query synthesis—and compare them. We apply the proposed framework to the problem of learning the price-demand function, an application that is important in optimal product pricing and dynamic (or time-varying) pricing. In our experiments, we provide a comparative study including the proposed framework strategies and some other baselines. The accomplished results demonstrate a significant performance for the proposed methods.

Highlights

Active learning has received a substantial growing interest in literature
All of the proposed active learning strategies including: our proposed strategies and the baseline methods, we implement two variants: one in pool-based setting and the other using query synthesis. In addition to these two active learning schemes, we further apply the third method, query synthesis without a predefined pool described in Section 5.1, to our proposed active learning methods that require the existence of a pool of unlabeled samples such as mutual information (MI), modified mutual Information (MMI) and Kullback–Leibler divergence (KL), in addition to the probabilistic-based exploration-exploitation (PEE) methods using either of MI, MMI or KL strategies for exploration
We propose a novel active learning framework for optimizing a general utility function

Summary

Introduction

Active learning has received a substantial growing interest in literature. Active learning is used for selecting the most informative “beneficial” training samples for the learning model in order to achieve high model accuracy using as few examples as possible [1]. Active learning has proved its superiority in diverse applications such as natural language processing [2] and image processing [3]. The active learning process basically proceeds as follows: first, an initial learning model is trained using a few training samples. Additional samples are sequentially added to the training data according to a certain querying strategy. This process repeats until a certain stopping criterion is satisfied [4]

Objectives

Methods

Results

Discussion

Conclusion