Causal Interpretation of Neural Network Computations with Contribution Decomposition

• URL: http://arxiv.org/abs/2603.06557v1
• Date: Fri, 06 Mar 2026 18:46:06 GMT
• Title: Causal Interpretation of Neural Network Computations with Contribution Decomposition
• Abstract summary: We take a direct approach to examine how hidden neurons act to drive network outputs.<n>We introduce CODEC, a method that uses sparse autoencoders to decompose network behavior into sparse motifs of hidden-neuron contributions.
• Score: 13.992892699439023
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Understanding how neural networks transform inputs into outputs is crucial for interpreting and manipulating their behavior. Most existing approaches analyze internal representations by identifying hidden-layer activation patterns correlated with human-interpretable concepts. Here we take a direct approach to examine how hidden neurons act to drive network outputs. We introduce CODEC (Contribution Decomposition), a method that uses sparse autoencoders to decompose network behavior into sparse motifs of hidden-neuron contributions, revealing causal processes that cannot be determined by analyzing activations alone. Applying CODEC to benchmark image-classification networks, we find that contributions grow in sparsity and dimensionality across layers and, unexpectedly, that they progressively decorrelate positive and negative effects on network outputs. We further show that decomposing contributions into sparse modes enables greater control and interpretation of intermediate layers, supporting both causal manipulations of network output and human-interpretable visualizations of distinct image components that combine to drive that output. Finally, by analyzing state-of-the-art models of neural activity in the vertebrate retina, we demonstrate that CODEC uncovers combinatorial actions of model interneurons and identifies the sources of dynamic receptive fields. Overall, CODEC provides a rich and interpretable framework for understanding how nonlinear computations evolve across hierarchical layers, establishing contribution modes as an informative unit of analysis for mechanistic insights into artificial neural networks.

Related papers

• Uncovering Semantic Selectivity of Latent Groups in Higher Visual Cortex with Mutual Information-Guided Diffusion [19.983291706164923]
  We present MIG-Vis, a method to visualize and validate the visual-semantic attributes encoded in neural latent subspaces.<n>We validate MIG-Vis on multi-session neural spiking datasets from the inferior temporal (IT) cortex of two macaques.
  arXiv  Detail & Related papers  (2025-10-02T16:33:40Z)
• Representation Understanding via Activation Maximization [13.88866465448849]
  We propose a unified feature visualization framework applicable to both Convolutional Neural Networks (CNNs) and Vision Transformers (ViTs)<n>Our experiments demonstrate the effectiveness of our approach in both traditional CNNs and modern ViT, highlighting its generalizability and value.
  arXiv  Detail & Related papers  (2025-08-10T10:36:30Z)
• Concept-Guided Interpretability via Neural Chunking [64.6429903327095]
  We show that neural networks exhibit patterns in their raw population activity that mirror regularities in the training data.<n>We propose three methods to extract recurring chunks on a neural population level.<n>Our work points to a new direction for interpretability, one that harnesses both cognitive principles and the structure of naturalistic data.
  arXiv  Detail & Related papers  (2025-05-16T13:49:43Z)
• Discovering Chunks in Neural Embeddings for Interpretability [53.80157905839065]
  We propose leveraging the principle of chunking to interpret artificial neural population activities.<n>We first demonstrate this concept in recurrent neural networks (RNNs) trained on artificial sequences with imposed regularities.<n>We identify similar recurring embedding states corresponding to concepts in the input, with perturbations to these states activating or inhibiting the associated concepts.
  arXiv  Detail & Related papers  (2025-02-03T20:30:46Z)
• Neural Networks Decoded: Targeted and Robust Analysis of Neural Network Decisions via Causal Explanations and Reasoning [9.947555560412397]
  We introduce TRACER, a novel method grounded in causal inference theory to estimate the causal dynamics underpinning DNN decisions. Our approach systematically intervenes on input features to observe how specific changes propagate through the network, affecting internal activations and final outputs. TRACER further enhances explainability by generating counterfactuals that reveal possible model biases and offer contrastive explanations for misclassifications.
  arXiv  Detail & Related papers  (2024-10-07T20:44:53Z)
• Automatic Discovery of Visual Circuits [66.99553804855931]
  We explore scalable methods for extracting the subgraph of a vision model's computational graph that underlies recognition of a specific visual concept. We find that our approach extracts circuits that causally affect model output, and that editing these circuits can defend large pretrained models from adversarial attacks.
  arXiv  Detail & Related papers  (2024-04-22T17:00:57Z)
• Graph Neural Networks for Learning Equivariant Representations of Neural Networks [55.04145324152541]
  We propose to represent neural networks as computational graphs of parameters. Our approach enables a single model to encode neural computational graphs with diverse architectures. We showcase the effectiveness of our method on a wide range of tasks, including classification and editing of implicit neural representations.
  arXiv  Detail & Related papers  (2024-03-18T18:01:01Z)
• Manipulating Feature Visualizations with Gradient Slingshots [53.94925202421929]
  Feature Visualization (FV) is a widely used technique for interpreting the concepts learned by Deep Neural Networks (DNNs)<n>We introduce a novel method, Gradient Slingshots, that enables manipulation of FV without modifying the model architecture or significantly degrading its performance.
  arXiv  Detail & Related papers  (2024-01-11T18:57:17Z)
• Neural Activation Patterns (NAPs): Visual Explainability of Learned Concepts [8.562628320010035]
  We present a method that takes into account the entire activation distribution. By extracting similar activation profiles within the high-dimensional activation space of a neural network layer, we find groups of inputs that are treated similarly. These input groups represent neural activation patterns (NAPs) and can be used to visualize and interpret learned layer concepts.
  arXiv  Detail & Related papers  (2022-06-20T09:05:57Z)
• Data-driven emergence of convolutional structure in neural networks [83.4920717252233]
  We show how fully-connected neural networks solving a discrimination task can learn a convolutional structure directly from their inputs. By carefully designing data models, we show that the emergence of this pattern is triggered by the non-Gaussian, higher-order local structure of the inputs.
  arXiv  Detail & Related papers  (2022-02-01T17:11:13Z)
• Dynamic Inference with Neural Interpreters [72.90231306252007]
  We present Neural Interpreters, an architecture that factorizes inference in a self-attention network as a system of modules. inputs to the model are routed through a sequence of functions in a way that is end-to-end learned. We show that Neural Interpreters perform on par with the vision transformer using fewer parameters, while being transferrable to a new task in a sample efficient manner.
  arXiv  Detail & Related papers  (2021-10-12T23:22:45Z)

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.