We propose to explain the predictions of a deep neural network, by pointing to the set of what we call representer points in the training set, for a given test point prediction. Specifically, we show that we can decompose the pre-activation prediction of a neural network into a linear combination of activations of training points, with the weights corresponding to what we call representer values, which thus capture the importance of that training point on the learned parameters of the network. But it provides a deeper understanding of the network than simply training point influence: with positive representer values corresponding to excitatory training points, and negative values corresponding to inhibitory points, which as we show provides considerably more insight. Our method is also much more scalable, allowing for real-time feedback in a manner not feasible with influence functions.
Representer Point Selection for Explaining Regularized High-dimensional Models
Carnegie Mellon University（卡内基梅隆大学）；Amazon Search（亚马逊搜索）；University of Illinois Urbana-Champaign（伊利诺伊大学厄巴纳-香槟分校）；University of California, Los Angeles（加利福尼亚大学洛杉矶分校）
How can we identify the training examples that contribute most to the prediction of a tree ensemble? In this paper, we introduce TREX, an explanation system that provides instance-attribution explanations for tree ensembles, such as random forests and gradient boosted trees. TREX builds on the representer point framework previously developed for explaining deep neural networks. Since tree ensembles are non-differentiable, we define a kernel that captures the structure of the specific tree ensemble. By using this kernel in kernel logistic regression or a support vector machine, TREX builds a su…
Modern machine learning systems based on neural networks have shown great success in learning complex data patterns while being able to make good predictions on unseen data points. However, the limited interpretability of these systems hinders further progress and application to several domains in the real world. This predicament is exemplified by time consuming model selection and the difficulties faced in predictive explainability, especially in the presence of adversarial examples. In this paper, we take a step towards better understanding of neural networks by introducing a local polytope …
NTT Communication Science Laboratories（NTT通信科学实验室）；Software Technology and Artificial Intelligence Research Laboratory（软件技术与人工智能研究实验室）；Chiba Institute of Technology（千叶工业大学）
Although deep models achieve high predictive performance, it is difficult for humans to understand the predictions they made. Explainability is important for real-world applications to justify their reliability. Many example-based explanation methods have been proposed, such as representer point selection, where an explanation model defined by a set of training examples is used for explaining a prediction model. For improving the interpretability, reducing the number of examples in the explanation model is important. However, the explanations with fewer examples can be unfaithful since it is d…
Predicting and Explaining Traffic Crash Severity Through Crash Feature Selection
Honda Research Institute Europe（霍恩达欧洲研究机构）；The Ohio State University（俄亥俄州立大学）；Department of Computer Science, University of Crete（克里特大学计算机科学系）；American Honda Motor Co., Inc.（美国本田摩托公司）
Identifying measurable genetic indicators (or biomarkers) of a specific condition of a biological system is a key element of precision medicine. Indeed it allows to tailor diagnostic, prognostic and treatment choice to individual characteristics of a patient. In machine learning terms, biomarker discovery can be framed as a feature selection problem on whole-genome data sets. However, classical feature selection methods are usually underpowered to process these data sets, which contain orders of magnitude more features than samples. This can be addressed by making the assumption that genetic f…
Graphs are ubiquitous in many applications, such as social networks, knowledge graphs, smart grids, etc.. Graph neural networks (GNN) are the current state-of-the-art for these applications, and yet remain obscure to humans. Explaining the GNN predictions can add transparency. However, as many graphs are not static but continuously evolving, explaining changes in predictions between two graph snapshots is different but equally important. Prior methods only explain static predictions or generate coarse or irrelevant explanations for dynamic predictions. We define the problem of explaining evolv…
University of Oxford（牛津大学）；Imperial College London（伦敦帝国理工学院）
We take a Bayesian perspective to illustrate a connection between training speed and the marginal likelihood in linear models. This provides two major insights: first, that a measure of a model's training speed can be used to estimate its marginal likelihood. Second, that this measure, under certain conditions, predicts the relative weighting of models in linear model combinations trained to minimize a regression loss. We verify our results in model selection tasks for linear models and for the infinite-width limit of deep neural networks. We further provide encouraging empirical evidence that…
Fraunhofer Heinrich Hertz Institute（弗劳恩霍夫海因里希·赫兹研究所）；BIFOLD – Berlin Institute for the Foundations of Learning and Data（BIFOLD——柏林学习与数据基础研究所）；Technische Universität Berlin（柏林工业大学）；Korea University（高丽大学）；Max Planck Institute for Informatics（马克斯·普朗克信息学研究所）
With the broader and highly successful usage of machine learning in industry and the sciences, there has been a growing demand for Explainable AI. Interpretability and explanation methods for gaining a better understanding about the problem solving abilities and strategies of nonlinear Machine Learning, in particular, deep neural networks, are therefore receiving increased attention. In this work we aim to (1) provide a timely overview of this active emerging field, with a focus on 'post-hoc' explanations, and explain its theoretical foundations, (2) put interpretability algorithms to a test b…
University of Campinas（坎皮纳斯大学）；Leeds University Business School（利兹大学商学院）；COMSATS University Islamabad（伊斯兰堡COMSATS大学）
The use of algorithm-agnostic approaches is an emerging area of research for explaining the contribution of individual features towards the predicted outcome. Whilst there is a focus on explaining the prediction itself, a little has been done on explaining the robustness of these models, that is, how each feature contributes towards achieving that robustness. In this paper, we propose the use of Shapley values to explain the contribution of each feature towards the model's robustness, measured in terms of Receiver-operating Characteristics (ROC) curve and the Area under the ROC curve (AUC). Wi…
CeADAR Irelands Centre for Applied AI, Technological University Dublin（爱尔兰CeADAR应用人工智能中心，都柏林理工大学）
Trust and credibility in machine learning models is bolstered by the ability of a model to explain itsdecisions. While explainability of deep learning models is a well-known challenge, a further chal-lenge is clarity of the explanation itself, which must be interpreted by downstream users. Layer-wiseRelevance Propagation (LRP), an established explainability technique developed for deep models incomputer vision, provides intuitive human-readable heat maps of input images. We present the novelapplication of LRP for the first time with structured datasets using a deep neural network (1D-CNN),for …
University of Cambridge（剑桥大学）；University of California Los Angeles（加州大学洛杉矶分校）；The Alan Turing Institute（艾伦·图灵研究所）
How can we explain the predictions of a machine learning model? When the data is structured as a multivariate time series, this question induces additional difficulties such as the necessity for the explanation to embody the time dependency and the large number of inputs. To address these challenges, we propose dynamic masks (Dynamask). This method produces instance-wise importance scores for each feature at each time step by fitting a perturbation mask to the input sequence. In order to incorporate the time dependency of the data, Dynamask studies the effects of dynamic perturbation operators…
Compact Example-Based Explanations for Language Models
Faculty of Computer Science, University of Vienna（维也纳大学计算机科学学院）；UniVie Doctoral School Computer Science, University of Vienna（维也纳大学计算机科学博士学院）；Faculty of Philological and Cultural Studies, University of Vienna（维也纳大学语言文化研究学院）
Indian Institute of Technology, Kharagpur（印度理工学院卡拉格普尔分校）；Hewlett Packard Enterprise, Bangalore（慧与（Hewlett Packard Enterprise）班加罗尔）
Finding valuable training data points for deep neural networks has been a core research challenge with many applications. In recent years, various techniques for calculating the "value" of individual training datapoints have been proposed for explaining trained models. However, the value of a training datapoint also depends on other selected training datapoints - a notion that is not explicitly captured by existing methods. In this paper, we study the problem of selecting high-value subsets of training data. The key idea is to design a learnable framework for online subset selection, which can…
University of Oxford（牛津大学）；Facebook AI Research（Facebook人工智能研究院）；Alan Turing Institute（艾伦·图灵研究所）；DeepMind
For AI systems to garner widespread public acceptance, we must develop methods capable of explaining the decisions of black-box models such as neural networks. In this work, we identify two issues of current explanatory methods. First, we show that two prevalent perspectives on explanations --- feature-additivity and feature-selection --- lead to fundamentally different instance-wise explanations. In the literature, explainers from different perspectives are currently being directly compared, despite their distinct explanation goals. The second issue is that current post-hoc explainers are eit…
