Guaranteed Optimal Compositional Explanations for Neurons

• URL: http://arxiv.org/abs/2511.20934v1
• Date: Tue, 25 Nov 2025 23:50:22 GMT
• Title: Guaranteed Optimal Compositional Explanations for Neurons
• Authors: Biagio La Rosa, Leilani H. Gilpin,
• Abstract summary: Compositional explanations describe the spatial alignment between neuron activations and concepts.<n> beam search cannot provide any theoretical guarantees of optimality.<n>We introduce the first framework for computing guaranteed optimal compositional explanations.
• Score: 4.497600020881818
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: While neurons are the basic units of deep neural networks, it is still unclear what they learn and if their knowledge is aligned with that of humans. Compositional explanations aim to answer this question by describing the spatial alignment between neuron activations and concepts through logical rules. These logical descriptions are typically computed via a search over all possible concept combinations. Since computing the spatial alignment over the entire state space is computationally infeasible, the literature commonly adopts beam search to restrict the space. However, beam search cannot provide any theoretical guarantees of optimality, and it remains unclear how close current explanations are to the true optimum. In this theoretical paper, we address this gap by introducing the first framework for computing guaranteed optimal compositional explanations. Specifically, we propose: (i) a decomposition that identifies the factors influencing the spatial alignment, (ii) a heuristic to estimate the alignment at any stage of the search, and (iii) the first algorithm that can compute optimal compositional explanations within a feasible time. Using this framework, we analyze the differences between optimal and non-optimal explanations in the most popular settings for compositional explanations, the computer vision domain and Convolutional Neural Networks. In these settings, we demonstrate that 10-40 percent of explanations obtained with beam search are suboptimal when overlapping concepts are involved. Finally, we evaluate a beam-search variant guided by our proposed decomposition and heuristic, showing that it matches or improves runtime over prior methods while offering greater flexibility in hyperparameters and computational resources.

Related papers

• Understanding Inverse Reinforcement Learning under Overparameterization: Non-Asymptotic Analysis and Global Optimality [52.906438147288256]
  We show that our algorithm can identify the globally optimal reward and policy under certain neural network structures.<n>This is the first IRL algorithm with a non-asymptotic convergence guarantee that provably achieves global optimality.
  arXiv  Detail & Related papers  (2025-03-22T21:16:08Z)
• CF-OPT: Counterfactual Explanations for Structured Prediction [47.36059095502583]
  optimization layers in deep neural networks have enjoyed a growing popularity in structured learning, improving the state of the art on a variety of applications. Yet, these pipelines lack interpretability since they are made of two opaque layers: a highly non-linear prediction model, such as a deep neural network, and an optimization layer, which is typically a complex black-box solver. Our goal is to improve the transparency of such methods by providing counterfactual explanations.
  arXiv  Detail & Related papers  (2024-05-28T15:48:27Z)
• Discrete Neural Algorithmic Reasoning [18.497863598167257]
  We propose to force neural reasoners to maintain the execution trajectory as a combination of finite predefined states.<n> trained with supervision on the algorithm's state transitions, such models are able to perfectly align with the original algorithm.
  arXiv  Detail & Related papers  (2024-02-18T16:03:04Z)
• Towards a fuller understanding of neurons with Clustered Compositional Explanations [8.440673378588489]
  We propose a generalization, called Clustered Compositional Explanations, that combines Compositional Explanations with clustering and a novel search to approximate a broader spectrum of the neurons' behavior. We define and address the problems connected to the application of these methods to multiple ranges of activations, analyze the insights retrievable by using our algorithm, and propose desiderata qualities that can be used to study the explanations returned by different algorithms.
  arXiv  Detail & Related papers  (2023-10-27T19:39:50Z)
• Unified Field Theory for Deep and Recurrent Neural Networks [56.735884560668985]
  We present a unified and systematic derivation of the mean-field theory for both recurrent and deep networks. We find that convergence towards the mean-field theory is typically slower for recurrent networks than for deep networks. Our method exposes that Gaussian processes are but the lowest order of a systematic expansion in $1/n$.
  arXiv  Detail & Related papers  (2021-12-10T15:06:11Z)
• Optimal Counterfactual Explanations in Tree Ensembles [3.8073142980733]
  We advocate for a model-based search aiming at "optimal" explanations and propose efficient mixed-integer programming approaches. We show that isolation forests can be modeled within our framework to focus the search on plausible explanations with a low outlier score.
  arXiv  Detail & Related papers  (2021-06-11T22:44:27Z)
• Lower Bounds and Optimal Algorithms for Smooth and Strongly Convex Decentralized Optimization Over Time-Varying Networks [79.16773494166644]
  We consider the task of minimizing the sum of smooth and strongly convex functions stored in a decentralized manner across the nodes of a communication network. We design two optimal algorithms that attain these lower bounds. We corroborate the theoretical efficiency of these algorithms by performing an experimental comparison with existing state-of-the-art methods.
  arXiv  Detail & Related papers  (2021-06-08T15:54:44Z)
• Learning Structures for Deep Neural Networks [99.8331363309895]
  We propose to adopt the efficient coding principle, rooted in information theory and developed in computational neuroscience. We show that sparse coding can effectively maximize the entropy of the output signals. Our experiments on a public image classification dataset demonstrate that using the structure learned from scratch by our proposed algorithm, one can achieve a classification accuracy comparable to the best expert-designed structure.
  arXiv  Detail & Related papers  (2021-05-27T12:27:24Z)
• On Guaranteed Optimal Robust Explanations for NLP Models [16.358394218953833]
  We build on abduction-based explanations for ma-chine learning and develop a method for computing local explanations for neural network models. We present two solution algorithms, respectively based on implicit hitting sets and maximum universal subsets. We evaluate our framework on three widely used sentiment analysis tasks and texts of up to100words from SST, Twitter and IMDB datasets.
  arXiv  Detail & Related papers  (2021-05-08T08:44:48Z)
• Sequential Subspace Search for Functional Bayesian Optimization Incorporating Experimenter Intuition [63.011641517977644]
  Our algorithm generates a sequence of finite-dimensional random subspaces of functional space spanned by a set of draws from the experimenter's Gaussian Process. Standard Bayesian optimisation is applied on each subspace, and the best solution found used as a starting point (origin) for the next subspace. We test our algorithm in simulated and real-world experiments, namely blind function matching, finding the optimal precipitation-strengthening function for an aluminium alloy, and learning rate schedule optimisation for deep networks.
  arXiv  Detail & Related papers  (2020-09-08T06:54:11Z)
• Parallelization Techniques for Verifying Neural Networks [52.917845265248744]
  We introduce an algorithm based on the verification problem in an iterative manner and explore two partitioning strategies. We also introduce a highly parallelizable pre-processing algorithm that uses the neuron activation phases to simplify the neural network verification problems.
  arXiv  Detail & Related papers  (2020-04-17T20:21:47Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.