Machine learning models that first learn a representation of a domain in terms of human-understandable concepts, then use it to make predictions, have been proposed to facilitate interpretation and interaction with models trained on high-dimensional data. However these methods have important limitations: the way they define concepts are not inherently interpretable, and they assume that concept labels either exist for individual instances or can easily be acquired from users. These limitations are particularly acute for high-dimensional tabular features. We propose an approach for learning a set of transparent concept definitions in high-dimensional tabular data that relies on users labeling concept features instead of individual instances. Our method produces concepts that both align with users' intuitive sense of what a concept means, and facilitate prediction of the downstream label by a transparent machine learning model. This ensures that the full model is transparent and intuitive, and as predictive as possible given this constraint. We demonstrate with simulated user feedback on real prediction problems, including one in a clinical domain, that this kind of direct feedback is much more efficient at learning solutions that align with ground truth concept definitions than alternative transparent approaches that rely on labeling instances or other existing interaction mechanisms, while maintaining similar predictive performance.
In this paper we propose and study the novel problem of explaining node embeddings by finding embedded human interpretable subspaces in already trained unsupervised node representation embeddings. We use an external knowledge base that is organized as a taxonomy of human-understandable concepts over entities as a guide to identify subspaces in node embeddings learned from an entity graph derived from Wikipedia. We propose a method that given a concept finds a linear transformation to a subspace where the structure of the concept is retained. Our initial experiments show that we obtain low erro…
In an attempt to gather a deeper understanding of how convolutional neural networks (CNNs) reason about human-understandable concepts, we present a method to infer labeled concept data from hidden layer activations and interpret the concepts through a shallow decision tree. The decision tree can provide information about which concepts a model deems important, as well as provide an understanding of how the concepts interact with each other. Experiments demonstrate that the extracted decision tree is capable of accurately representing the original CNN's classifications at low tree depths, thus …
AXA（安盛集团）；OECD（经济合作与发展组织）；Sorbonne Universite（索邦大学）；CNRS（法国国家科学研究中心）；Polish Academy of Science（波兰科学院）
Interpretable surrogates of black-box predictors trained on high-dimensional tabular datasets can struggle to generate comprehensible explanations in the presence of correlated variables. We propose a model-agnostic interpretable surrogate that provides global and local explanations of black-box classifiers to address this issue. We introduce the idea of concepts as intuitive groupings of variables that are either defined by a domain expert or automatically discovered using correlation coefficients. Concepts are embedded in a surrogate decision tree to enhance its comprehensibility. First expe…
Technische Universität Berlin（柏林工业大学）；Max Planck Institute for Human Cognitive and Brain Sciences（马克斯·普朗克人类认知与脑科学研究所）；Berlin Institute for the Foundations of Learning and Data（柏林学习与数据基础研究所）；National Institute of Mental Health（国家精神卫生研究所）；Naval Postgraduate School（海军研究生院）
A central goal in the cognitive sciences is the development of numerical models for mental representations of object concepts. This paper introduces Variational Interpretable Concept Embeddings (VICE), an approximate Bayesian method for embedding object concepts in a vector space using data collected from humans in a triplet odd-one-out task. VICE uses variational inference to obtain sparse, non-negative representations of object concepts with uncertainty estimates for the embedding values. These estimates are used to automatically select the dimensions that best explain the data. We derive a …
Few-shot learning aims at recognizing new instances from classes with limited samples. This challenging task is usually alleviated by performing meta-learning on similar tasks. However, the resulting models are black-boxes. There has been growing concerns about deploying black-box machine learning models and FSL is not an exception in this regard. In this paper, we propose a method for FSL based on a set of human-interpretable concepts. It constructs a set of metric spaces associated with the concepts and classifies samples of novel classes by aggregating concept-specific decisions. The propos…
Interpretable Reward Modeling with Active Concept Bottlenecks
Department of Computer Science, ETH Zurich, Zurich, Switzerland（苏黎世联邦理工学院计算机科学系）；Department of Applied Mathematics and Theoretical Physics, University of Cambridge, United Kingdom（剑桥大学应用数学与理论物理系）
ETH Zürich（苏黎世联邦理工学院）；Max-Planck-Institute for Intelligent Systems（马克斯·普朗克智能系统研究所）
In safety-critical applications, practitioners are reluctant to trust neural networks when no interpretable explanations are available. Many attempts to provide such explanations revolve around pixel-based attributions or use previously known concepts. In this paper we aim to provide explanations by provably identifying \emph{high-level, previously unknown ground-truth concepts}. To this end, we propose a probabilistic modeling framework to derive (C)oncept (L)earning and (P)rediction (CLAP) -- a VAE-based classifier that uses visually interpretable concepts as predictors for a simple classifi…
University of Cambridge（剑桥大学）；Université Côte d’Azur（蔚蓝海岸大学）；CNRS（法国国家科学研究中心）；Inria（法国国家信息与自动化研究所）；University of Siena（锡耶纳大学）；INRA（法国国家农业食品与环境研究院）；Université Paris-Saclay（巴黎萨克雷大学）；KU Leuven（鲁汶大学）
Learning Interpretable Concepts: Unifying Causal Representation Learning and Foundation Models
Carnegie Mellon University（卡内基梅隆大学）；Max Planck Institute for Intelligent Systems（马克斯·普朗克智能系统研究所）；Tübingen AI Center（蒂宾根人工智能中心）；University of Chicago（芝加哥大学）；ELLIS Institute Tübingen（ELLIS蒂宾根研究所）
Tree of Concepts: Interpretable Continual Learners in Non-Stationary Clinical Domains
Computer Science Department（计算机科学系）；New York University（纽约大学）；Department of Population Health（人口健康部）；NYU Grossman School of Medicine（NYU Grossman医学院）；School of Global Public Health（全球公共卫生学院）
本文提出Tree of Concepts框架，通过浅层决策树定义固定规则概念接口，并训练概念瓶颈模型预测概念，实现持续学习中的可解释性和稳定性平衡。
Multi-agent robotic systems are increasingly operating in real-world environments in close proximity to humans, yet are largely controlled by policy models with inscrutable deep neural network representations. We introduce a method for incorporating interpretable concepts from a domain expert into models trained through multi-agent reinforcement learning, by requiring the model to first predict such concepts then utilize them for decision making. This allows an expert to both reason about the resulting concept policy models in terms of these high-level concepts at run-time, as well as interven…
Technical University of Kaiserslautern（凯泽斯劳滕工业大学）；Technical University of Berlin（柏林工业大学）；University of Birmingham（伯明翰大学）；NPZ Innovation GmbH（NPZ创新有限公司）；University of St.Gallen（圣加仑大学）
We propose a novel training methodology -- Concept Group Learning (CGL) -- that encourages training of interpretable CNN filters by partitioning filters in each layer into concept groups, each of which is trained to learn a single visual concept. We achieve this through a novel regularization strategy that forces filters in the same group to be active in similar image regions for a given layer. We additionally use a regularizer to encourage a sparse weighting of the concept groups in each layer so that a few concept groups can have greater importance than others. We quantitatively evaluate CGL…
York University（约克大学）；University of Toronto（多伦多大学）；Kempner Institute, Harvard University（哈佛大学凯姆纳研究所）；Vector Institute（向量研究所）；Samsung AI Centre（三星人工智能中心）；Trajectory Labs（轨迹实验室）
Restyling Unsupervised Concept Based Interpretable Networks with Generative Models
Sorbonne Université（索邦大学）；Technical University of Munich（慕尼黑工业大学）；Helmholtz Munich（亥姆霍兹慕尼黑中心）；Munich Center for Machine Learning (MCML)（慕尼黑机器学习中心）；LTCI, Télécom Paris, Institut Polytechnique de Paris（巴黎理工学院 巴黎高等电信学院 LTCI）
Unsupervised Interpretable Basis Extraction for Concept-Based Visual Explanations
Information Technologies Institute, Centre for Research and Technology HELLAS（信息科技研究所，希腊研究中心与技术HELLOS）；Department of Advanced Computing Sciences, University of Maastricht（高级计算科学系，马斯特里赫特大学）
University of Bristol（布里斯托大学）；CortAIx, Thales（泰雷兹公司 CortAIx）
The potential of reinforcement learning (RL) to deliver aligned and performant agents is partially bottlenecked by the reward engineering problem. One alternative to heuristic trial-and-error is preference-based RL (PbRL), where a reward function is inferred from sparse human feedback. However, prior PbRL methods lack interpretability of the learned reward structure, which hampers the ability to assess robustness and alignment. We propose an online, active preference learning algorithm that constructs reward functions with the intrinsically interpretable, compositional structure of a tree. Usi…
CAT: Interpretable Concept-based Taylor Additive Models
William & Mary（威廉玛丽学院）；AT&T CDO（美国电话电报公司首席数字官办公室）；University of Illinois at Urbana-Champaign（伊利诺伊大学厄巴纳-香槟分校）；The University of North Carolina at Chapel Hill（北卡罗来纳大学教堂山分校）
Using attention weights to identify information that is important for models' decision-making is a popular approach to interpret attention-based neural networks. This is commonly realized in practice through the generation of a heat-map for every single document based on attention weights. However, this interpretation method is fragile, and easy to find contradictory examples. In this paper, we propose a corpus-level explanation approach, which aims to capture causal relationships between keywords and model predictions via learning the importance of keywords for predicted labels across a train…
What does a neural network encode about a concept as we traverse through the layers? Interpretability in machine learning is undoubtedly important, but the calculations of neural networks are very challenging to understand. Attempts to see inside their hidden layers can either be misleading, unusable, or rely on the latent space to possess properties that it may not have. In this work, rather than attempting to analyze a neural network posthoc, we introduce a mechanism, called concept whitening (CW), to alter a given layer of the network to allow us to better understand the computation leading…
Single-cell RNA-seq foundation models achieve strong performance on downstream tasks but remain black boxes, limiting their utility for biological discovery. Recent work has shown that sparse dictionary learning can extract concepts from deep learning models, with promising applications in biomedical imaging and protein models. However, interpreting biological concepts remains challenging, as biological sequences are not inherently human-interpretable. We introduce a novel concept-based interpretability framework for single-cell RNA-seq models with a focus on concept interpretation and evaluat…
Closing the Loop: PID Feedback Control for Interpretable Activation Steering in Symbolic Music Generation
Athens University of Economics and Business（雅典经济与商业大学）；Orfium Research（Orfium 研究）；Hellenic Mediterranean University（希腊地中海大学）；Archimedes / Athena Research Center（阿基米德/雅典娜研究中心）
CI-CBM: Class-Incremental Concept Bottleneck Model for Interpretable Continual Learning
Department of Electrical and Computer Engineering, University of California San Diego（加州大学圣地亚哥分校电气与计算机工程系）；Department of Computer Science and Engineering, University of California San Diego（加州大学圣地亚哥分校计算机科学与工程系）；Halıcıoğlu Data Science Institute, University of California San Diego（加州大学圣地亚哥分校Halıcıoğlu数据科学研究所）
Department of Computer Science and Technology, University of Cambridge, Cambridge, UK（计算机科学与技术系，剑桥大学，剑桥，英国）；Trinity College, University of Oxford, Oxford, UK（牛津大学三一学院，牛津，英国）；Cambridge Institute for Technology and Humanity, Cambridge, UK（剑桥技术与人类研究所，剑桥，英国）
When Confidence Lacks Concepts: Interpretable OOD Detection via Representation Perturbations
NepAl Applied Mathematics and Informatics Institute for research（尼泊尔应用数学与信息学研究所）；West Virginia University（西弗吉尼亚大学）；Kathmandu University（加德满都大学）；University College London（伦敦大学学院）；University of Aberdeen（阿伯丁大学）
