Feature Integration Spaces: Joint Training Reveals Dual Encoding in Neural Network Representations

• URL: http://arxiv.org/abs/2507.00269v1
• Date: Mon, 30 Jun 2025 21:26:58 GMT
• Title: Feature Integration Spaces: Joint Training Reveals Dual Encoding in Neural Network Representations
• Authors: Omar Claflin,
• Abstract summary: Current autoencoder (SAE) approaches to neural network interpretability assume that activations can be decomposed through linear superposition into sparse, interpretable features.<n>We propose that neural networks encode information in two complementary spaces compressed into the same substrate: feature identity and feature integration.<n>Joint training achieves 41.3% reconstruction improvement and 51.6% reduction in KL divergence errors.
• Score: 0.0
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Current sparse autoencoder (SAE) approaches to neural network interpretability assume that activations can be decomposed through linear superposition into sparse, interpretable features. Despite high reconstruction fidelity, SAEs consistently fail to eliminate polysemanticity and exhibit pathological behavioral errors. We propose that neural networks encode information in two complementary spaces compressed into the same substrate: feature identity and feature integration. To test this dual encoding hypothesis, we develop sequential and joint-training architectures to capture identity and integration patterns simultaneously. Joint training achieves 41.3% reconstruction improvement and 51.6% reduction in KL divergence errors. This architecture spontaneously develops bimodal feature organization: low squared norm features contributing to integration pathways and the rest contributing directly to the residual. Small nonlinear components (3% of parameters) achieve 16.5% standalone improvements, demonstrating parameter-efficient capture of computational relationships crucial for behavior. Additionally, intervention experiments using 2x2 factorial stimulus designs demonstrated that integration features exhibit selective sensitivity to experimental manipulations and produce systematic behavioral effects on model outputs, including significant interaction effects across semantic dimensions. This work provides systematic evidence for (1) dual encoding in neural representations, (2) meaningful nonlinearly encoded feature interactions, and (3) introduces an architectural paradigm shift from post-hoc feature analysis to integrated computational design, establishing foundations for next-generation SAEs.

Related papers

• StdGEN++: A Comprehensive System for Semantic-Decomposed 3D Character Generation [57.06461272772509]
  StdGEN++ is a novel and comprehensive system for generating high-fidelity, semantically decomposed 3D characters from diverse inputs.<n>It achieves state-of-the-art performance, significantly outperforming existing methods in geometric accuracy and semantic disentanglement.<n>The resulting structural independence unlocks advanced downstream capabilities, including non-destructive editing, physics-compliant animation, and gaze tracking.
  arXiv  Detail & Related papers  (2026-01-12T15:41:27Z)
• STORE: Semantic Tokenization, Orthogonal Rotation and Efficient Attention for Scaling Up Ranking Models [11.965535230928372]
  Store is a unified and scalable token-based ranking framework built upon three core innovations.<n>Our framework consistently improves prediction accuracy(online CTR by 2.71%, AUC by 1.195%) and training effeciency (1.84 throughput)
  arXiv  Detail & Related papers  (2025-11-24T06:20:02Z)
• Knowledge-Informed Neural Network for Complex-Valued SAR Image Recognition [51.03674130115878]
  We introduce the Knowledge-Informed Neural Network (KINN), a lightweight framework built upon a novel "compression-aggregation-compression" architecture.<n>KINN establishes a state-of-the-art in parameter-efficient recognition, offering exceptional generalization in data-scarce and out-of-distribution scenarios.
  arXiv  Detail & Related papers  (2025-10-23T07:12:26Z)
• Bidirectional Representations Augmented Autoregressive Biological Sequence Generation:Application in De Novo Peptide Sequencing [51.12821379640881]
  Autoregressive (AR) models offer holistic, bidirectional representations but face challenges with generative coherence and scalability.<n>We propose a hybrid framework enhancing AR generation by dynamically integrating rich contextual information from non-autoregressive mechanisms.<n>A novel cross-decoder attention module enables the AR decoder to iteratively query and integrate these bidirectional features.
  arXiv  Detail & Related papers  (2025-10-09T12:52:55Z)
• A Unified Cortical Circuit Model with Divisive Normalization and Self-Excitation for Robust Representation and Memory Maintenance [2.705743343600661]
  We introduce a recurrent neural circuit that combines divisive normalization with self-excitation to achieve robust encoding.<n>We demonstrate the model's versatility in two canonical tasks.<n>This work establishes a unified mathematical framework that bridges noise suppression, working memory, and approximate Bayesian inference.
  arXiv  Detail & Related papers  (2025-08-18T08:00:24Z)
• Uncovering Magnetic Phases with Synthetic Data and Physics-Informed Training [0.0]
  We investigate the efficient learning of magnetic phases using artificial neural networks trained on synthetic data.<n>We incorporate two key forms of physics-informed guidance to enhance model performance.<n>Our results show that synthetic, structured, and computationally efficient training schemes can reveal physically meaningful phase boundaries.
  arXiv  Detail & Related papers  (2025-05-15T15:16:16Z)
• Hybrid Two-Stage Reconstruction of Multiscale Subsurface Flow with Physics-informed Residual Connected Neural Operator [4.303037819686676]
  We propose a hybrid two-stage framework that uses multiscale basis functions and physics-guided deep learning to solve the Darcy flow problem.<n>The framework achieves R2 values above 0.9 in terms of basis function fitting and pressure reconstruction, and the residual indicator is on the order of $1times 10-4$.
  arXiv  Detail & Related papers  (2025-01-22T23:28:03Z)
• TokenUnify: Scaling Up Autoregressive Pretraining for Neuron Segmentation [65.65530016765615]
  We propose a hierarchical predictive coding framework that captures multi-scale dependencies through three complementary learning objectives.<n> TokenUnify integrates random token prediction, next-token prediction, and next-all token prediction to create a comprehensive representational space.<n>We also introduce a large-scale EM dataset with 1.2 billion annotated voxels, offering ideal long-sequence visual data with spatial continuity.
  arXiv  Detail & Related papers  (2024-05-27T05:45:51Z)
• SNE-RoadSegV2: Advancing Heterogeneous Feature Fusion and Fallibility Awareness for Freespace Detection [29.348921424716057]
  This paper presents a novel heterogeneous feature fusion block, comprising a holistic attention module, a heterogeneous feature contrast descriptor, and an affinity-weighted feature recalibrator. It incorporates both inter-scale and intra-scale skip connections into the decoder architecture while eliminating redundant ones, leading to both improved accuracy and computational efficiency. It introduces two fallibility-aware loss functions that separately focus on semantic-transition and depth-inconsistent regions, collectively contributing to greater supervision during model training.
  arXiv  Detail & Related papers  (2024-02-29T07:20:02Z)
• Adaptive Activation Functions for Predictive Modeling with Sparse Experimental Data [2.012425476229879]
  This study investigates the influence of adaptive or trainable activation functions on classification accuracy and predictive uncertainty in settings characterized by limited data availability. Our investigation reveals that adaptive activation functions, such as Exponential Linear Unit (ELU) and Softplus, with individual trainable parameters, result in accurate and confident prediction models.
  arXiv  Detail & Related papers  (2024-02-08T04:35:09Z)
• Fully Differentiable Correlation-driven 2D/3D Registration for X-ray to CT Image Fusion [3.868072865207522]
  Image-based rigid 2D/3D registration is a critical technique for fluoroscopic guided surgical interventions. We propose a novel fully differentiable correlation-driven network using a dual-branch CNN-transformer encoder. A correlation-driven loss is proposed for low-frequency feature and high-frequency feature decomposition based on embedded information.
  arXiv  Detail & Related papers  (2024-02-04T14:12:51Z)
• Understanding Augmentation-based Self-Supervised Representation Learning via RKHS Approximation and Regression [53.15502562048627]
  Recent work has built the connection between self-supervised learning and the approximation of the top eigenspace of a graph Laplacian operator. This work delves into a statistical analysis of augmentation-based pretraining.
  arXiv  Detail & Related papers  (2023-06-01T15:18:55Z)
• Improving Neural Additive Models with Bayesian Principles [54.29602161803093]
  Neural additive models (NAMs) enhance the transparency of deep neural networks by handling calibrated input features in separate additive sub-networks. We develop Laplace-approximated NAMs (LA-NAMs) which show improved empirical performance on datasets and challenging real-world medical tasks.
  arXiv  Detail & Related papers  (2023-05-26T13:19:15Z)
• Self-Supervised Generative-Contrastive Learning of Multi-Modal Euclidean Input for 3D Shape Latent Representations: A Dynamic Switching Approach [53.376029341079054]
  We propose a combined generative and contrastive neural architecture for learning latent representations of 3D shapes.<n>The architecture uses two encoder branches for voxel grids and multi-view images from the same underlying shape.
  arXiv  Detail & Related papers  (2023-01-11T18:14:24Z)
• A Generic Shared Attention Mechanism for Various Backbone Neural Networks [53.36677373145012]
  Self-attention modules (SAMs) produce strongly correlated attention maps across different layers. Dense-and-Implicit Attention (DIA) shares SAMs across layers and employs a long short-term memory module. Our simple yet effective DIA can consistently enhance various network backbones.
  arXiv  Detail & Related papers  (2022-10-27T13:24:08Z)
• 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)
• CSformer: Bridging Convolution and Transformer for Compressive Sensing [65.22377493627687]
  This paper proposes a hybrid framework that integrates the advantages of leveraging detailed spatial information from CNN and the global context provided by transformer for enhanced representation learning. The proposed approach is an end-to-end compressive image sensing method, composed of adaptive sampling and recovery. The experimental results demonstrate the effectiveness of the dedicated transformer-based architecture for compressive sensing.
  arXiv  Detail & Related papers  (2021-12-31T04:37:11Z)
• Spatio-Temporal Representation Factorization for Video-based Person Re-Identification [55.01276167336187]
  We propose Spatio-Temporal Representation Factorization module (STRF) for re-ID. STRF is a flexible new computational unit that can be used in conjunction with most existing 3D convolutional neural network architectures for re-ID. We empirically show that STRF improves performance of various existing baseline architectures while demonstrating new state-of-the-art results.
  arXiv  Detail & Related papers  (2021-07-25T19:29:37Z)
• Relaxing the Constraints on Predictive Coding Models [62.997667081978825]
  Predictive coding is an influential theory of cortical function which posits that the principal computation the brain performs is the minimization of prediction errors. Standard implementations of the algorithm still involve potentially neurally implausible features such as identical forward and backward weights, backward nonlinear derivatives, and 1-1 error unit connectivity. In this paper, we show that these features are not integral to the algorithm and can be removed either directly or through learning additional sets of parameters with Hebbian update rules without noticeable harm to learning performance.
  arXiv  Detail & Related papers  (2020-10-02T15:21:37Z)

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.