{"title":"ANN 分区单元的逻辑解释","authors":"Ingo Schmitt","doi":"arxiv-2408.14314","DOIUrl":null,"url":null,"abstract":"Consider a binary classification problem solved using a feed-forward\nartificial neural network (ANN). Let the ANN be composed of a ReLU layer and\nseveral linear layers (convolution, sum-pooling, or fully connected). We assume\nthe network was trained with high accuracy. Despite numerous suggested\napproaches, interpreting an artificial neural network remains challenging for\nhumans. For a new method of interpretation, we construct a bridge between a\nsimple ANN and logic. As a result, we can analyze and manipulate the semantics\nof an ANN using the powerful tool set of logic. To achieve this, we decompose\nthe input space of the ANN into several network partition cells. Each network\npartition cell represents a linear combination that maps input values to a\nclassifying output value. For interpreting the linear map of a partition cell\nusing logic expressions, we suggest minterm values as the input of a simple\nANN. We derive logic expressions representing interaction patterns for\nseparating objects classified as 1 from those classified as 0. To facilitate an\ninterpretation of logic expressions, we present them as binary logic trees.","PeriodicalId":501208,"journal":{"name":"arXiv - CS - Logic in Computer Science","volume":"395 1","pages":""},"PeriodicalIF":0.0000,"publicationDate":"2024-08-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Logic interpretations of ANN partition cells\",\"authors\":\"Ingo Schmitt\",\"doi\":\"arxiv-2408.14314\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"Consider a binary classification problem solved using a feed-forward\\nartificial neural network (ANN). Let the ANN be composed of a ReLU layer and\\nseveral linear layers (convolution, sum-pooling, or fully connected). We assume\\nthe network was trained with high accuracy. Despite numerous suggested\\napproaches, interpreting an artificial neural network remains challenging for\\nhumans. For a new method of interpretation, we construct a bridge between a\\nsimple ANN and logic. As a result, we can analyze and manipulate the semantics\\nof an ANN using the powerful tool set of logic. To achieve this, we decompose\\nthe input space of the ANN into several network partition cells. Each network\\npartition cell represents a linear combination that maps input values to a\\nclassifying output value. For interpreting the linear map of a partition cell\\nusing logic expressions, we suggest minterm values as the input of a simple\\nANN. We derive logic expressions representing interaction patterns for\\nseparating objects classified as 1 from those classified as 0. To facilitate an\\ninterpretation of logic expressions, we present them as binary logic trees.\",\"PeriodicalId\":501208,\"journal\":{\"name\":\"arXiv - CS - Logic in Computer Science\",\"volume\":\"395 1\",\"pages\":\"\"},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2024-08-26\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"arXiv - CS - Logic in Computer Science\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/arxiv-2408.14314\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"arXiv - CS - Logic in Computer Science","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/arxiv-2408.14314","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 0
摘要
考虑使用前馈人工神经网络(ANN)解决二元分类问题。让人工神经网络由一个 ReLU 层和多个线性层(卷积层、求和池层或全连接层)组成。我们假设该网络经过高精度训练。尽管提出了许多方法,但对于人类来说,解读人工神经网络仍然是一项挑战。作为一种新的解释方法,我们在简单的人工神经网络和逻辑之间架起了一座桥梁。因此,我们可以使用强大的逻辑工具集来分析和操纵人工神经网络的语义。为此,我们将 ANN 的输入空间分解为多个网络分区单元。每个网络分区单元代表一个将输入值映射到分类输出值的线性组合。为了用逻辑表达式解释分区单元的线性映射,我们建议将 minterm 值作为简单 ANN 的输入。为了便于解释逻辑表达式,我们将其表述为二进制逻辑树。
Consider a binary classification problem solved using a feed-forward
artificial neural network (ANN). Let the ANN be composed of a ReLU layer and
several linear layers (convolution, sum-pooling, or fully connected). We assume
the network was trained with high accuracy. Despite numerous suggested
approaches, interpreting an artificial neural network remains challenging for
humans. For a new method of interpretation, we construct a bridge between a
simple ANN and logic. As a result, we can analyze and manipulate the semantics
of an ANN using the powerful tool set of logic. To achieve this, we decompose
the input space of the ANN into several network partition cells. Each network
partition cell represents a linear combination that maps input values to a
classifying output value. For interpreting the linear map of a partition cell
using logic expressions, we suggest minterm values as the input of a simple
ANN. We derive logic expressions representing interaction patterns for
separating objects classified as 1 from those classified as 0. To facilitate an
interpretation of logic expressions, we present them as binary logic trees.