Evolutionary model building under streaming data for classification tasks: opportunities and challenges

Abstract

Streaming data analysis potentially represents a significant shift in emphasis from schemes historically pursued for offline (batch) approaches to the classification task. In particular, a streaming data application implies that: (1) the data itself has no formal 'start' or 'end'; (2) the properties of the process generating the data are non-stationary, thus models that function correctly for some part(s) of a stream may be ineffective elsewhere; (3) constraints on the time to produce a response, potentially implying an anytime operational requirement; and (4) given the prohibitive cost of employing an oracle to label a stream, a finite labelling budget is necessary. The scope of this article is to provide a survey of developments for model building under streaming environments from the perspective of both evolutionary and non-evolutionary frameworks. In doing so, we bring attention to the challenges and opportunities that developing solutions to streaming data classification tasks are likely to face using evolutionary approaches.

Evolving Hierarchical Memory-Prediction Machines in Multi-Task Reinforcement Learning

“…This is significant because temporal problem decomposition is likely beneficial in dynamic, non-stationary environments. Examples of this include MTRL, as well as time series forecasting or streaming data classification tasks when the underlying process generating the data stream changes significantly over time [1,15]. Putting these developments together, the overall purview of this work is to demonstrate how TPG can be used to build hierarchical memory-prediction machines that address the MTRL challenges outlined in Section 1.1.…”

Modelling Evolvability in Genetic Programming

“…Instead, current literature efforts to exploit evolvability do so indirectly, without having to measure it, such as defining new evolvability metrics [82] and characterizing evolvability's relatedness to other properties [35]. There has been some success in determining how much to select for evolvability, but only under limited circumstances [91,92].…”

“…Heywood [35] provides a survey of both evolutionary and non-evolutionary model developments for streaming data classification tasks. A series of works by Vahdat et al [84,85,86] use GP to classify streaming data with a variety of improvements to SGP which address problems that occur when dealing with streaming data.…”