Evolutionary Design of Reduced Precision Preprocessor for Levodopa-Induced Dyskinesia Classifier

Evolutionary Design of Reduced Precision Preprocessor for Levodopa-Induced Dyskinesia Classifier

Paper in proceedings (conference paper)

The aim of this work is to design a hardware-efficient implementation of data preprocessing in the task of levodopa-induced dyskinesia classification. In this task, there are three approaches implemented and compared: 1) evolution of magnitude approximation using Cartesian genetic programming, 2) design of preprocessing unit using two-population coevolution (2P-CoEA) of cartesian programs and fitness predictors, which are small subsets of training set, and 3) a design using three-population coevolution (3P-CoEA) combining compositional coevolution of preprocessor and classifier with coevolution of fitness predictors. Experimental results show that all of the three investigated approaches are capable of producing energy-saving solutions, suitable for implementation in hardware unit, with a quality comparable to baseline software implementation. Design of approximate magnitude leads to correctly working solutions, however, more energy-demanding than other investigated approaches. 3P-CoEA is capable of designing both preprocessor and classifier compositionally while achieving smaller solutions than the design of approximate magnitude. Presented 2P-CoEA results in the smallest and the most energy-efficient solutions along with producing a solution with significantly better classification quality for one part of test data in comparison with the software implementation.

The aim of this work is to design a hardware-efficient implementation of data preprocessing in the task of levodopa-induced dyskinesia classification. In this task, there are three approaches implemented and compared: 1) evolution of magnitude approximation using Cartesian genetic programming, 2) design of preprocessing unit using two-population coevolution (2P-CoEA) of cartesian programs and fitness predictors, which are small subsets of training set, and 3) a design using three-population coevolution (3P-CoEA) combining compositional coevolution of preprocessor and classifier with coevolution of fitness predictors. Experimental results show that all of the three investigated approaches are capable of producing energy-saving solutions, suitable for implementation in hardware unit, with a quality comparable to baseline software implementation. Design of approximate magnitude leads to correctly working solutions, however, more energy-demanding than other investigated approaches. 3P-CoEA is capable of designing both preprocessor and classifier compositionally while achieving smaller solutions than the design of approximate magnitude. Presented 2P-CoEA results in the smallest and the most energy-efficient solutions along with producing a solution with significantly better classification quality for one part of test data in comparison with the software implementation.

Cartesian genetic programming, compositional coevolution, adaptive size fitness predictors, levodopa-induced dyskinesia, approximate magnitude, energy-efficient

Cartesian genetic programming, compositional coevolution, adaptive size fitness predictors, levodopa-induced dyskinesia, approximate magnitude, energy-efficient

HURTA, M.; DRAHOŠOVÁ, M.; MRÁZEK, V.

2023

10.09.2022

Springer Nature Switzerland AG

Dortmund

978-3-031-14713-5

Parallel Problem Solving from Nature - PPSN XVII

Lecture Notes in Computer Science

13398

491

504

14

https://link.springer.com/chapter/10.1007/978-3-031-14714-2_34