Jorge Torres Gómez;Pit Hofmann;Frank H. P. Fitzek;Falko Dressler
{"title":"Explainability of Neural Networks for Symbol Detection in Molecular Communication Channels","authors":"Jorge Torres Gómez;Pit Hofmann;Frank H. P. Fitzek;Falko Dressler","doi":"10.1109/TMBMC.2023.3297135","DOIUrl":null,"url":null,"abstract":"Recent molecular communication (MC) research suggests machine learning (ML) models for symbol detection, avoiding the unfeasibility of end-to-end channel models. However, ML models are applied as black boxes, lacking proof of correctness of the underlying neural networks (NNs) to detect incoming symbols. This paper studies approaches to the explainability of NNs for symbol detection in MC channels. Based on MC channel models and real testbed measurements, we generate synthesized data and train a NN model to detect of binary transmissions in MC channels. Using the local interpretable model-agnostic explanation (LIME) method and the individual conditional expectation (ICE), the findings in this paper demonstrate the analogy between the trained NN and the standard peak and slope detectors.","PeriodicalId":36530,"journal":{"name":"IEEE Transactions on Molecular, Biological, and Multi-Scale Communications","volume":null,"pages":null},"PeriodicalIF":2.4000,"publicationDate":"2023-07-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"1","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"IEEE Transactions on Molecular, Biological, and Multi-Scale Communications","FirstCategoryId":"1085","ListUrlMain":"https://ieeexplore.ieee.org/document/10188873/","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q2","JCRName":"ENGINEERING, ELECTRICAL & ELECTRONIC","Score":null,"Total":0}
引用次数: 1
Abstract
Recent molecular communication (MC) research suggests machine learning (ML) models for symbol detection, avoiding the unfeasibility of end-to-end channel models. However, ML models are applied as black boxes, lacking proof of correctness of the underlying neural networks (NNs) to detect incoming symbols. This paper studies approaches to the explainability of NNs for symbol detection in MC channels. Based on MC channel models and real testbed measurements, we generate synthesized data and train a NN model to detect of binary transmissions in MC channels. Using the local interpretable model-agnostic explanation (LIME) method and the individual conditional expectation (ICE), the findings in this paper demonstrate the analogy between the trained NN and the standard peak and slope detectors.
期刊介绍:
As a result of recent advances in MEMS/NEMS and systems biology, as well as the emergence of synthetic bacteria and lab/process-on-a-chip techniques, it is now possible to design chemical “circuits”, custom organisms, micro/nanoscale swarms of devices, and a host of other new systems. This success opens up a new frontier for interdisciplinary communications techniques using chemistry, biology, and other principles that have not been considered in the communications literature. The IEEE Transactions on Molecular, Biological, and Multi-Scale Communications (T-MBMSC) is devoted to the principles, design, and analysis of communication systems that use physics beyond classical electromagnetism. This includes molecular, quantum, and other physical, chemical and biological techniques; as well as new communication techniques at small scales or across multiple scales (e.g., nano to micro to macro; note that strictly nanoscale systems, 1-100 nm, are outside the scope of this journal). Original research articles on one or more of the following topics are within scope: mathematical modeling, information/communication and network theoretic analysis, standardization and industrial applications, and analytical or experimental studies on communication processes or networks in biology. Contributions on related topics may also be considered for publication. Contributions from researchers outside the IEEE’s typical audience are encouraged.