Context-dependent manifold learning: A neuromodulated constrained autoencoder approach

2026-03-12

AI Summary

Standard constrained autoencoders struggle to adapt to varying physical parameters or environmental changes in non-stationary settings and often conflate primary inputs with contextual information. To address this, this work proposes the Neurally modulated Constrained Autoencoder (NcAE), which introduces neural modulation into the constrained autoencoding framework for the first time. By dynamically adjusting the gain and bias of geometric constraints through static contextual signals, NcAE enables context-aware manifold learning. This approach effectively disentangles global context from local manifold representations, preserving strict projection properties while accurately capturing the dynamic variations in manifold geometry across different environmental conditions. As a result, the model exhibits significantly enhanced adaptability and representational capacity in non-stationary physical systems.

Technology Category

• Machine Learning: Learning with Manifolds
• Constraint Satisfaction and Optimization: Constraint Learning and Acquisition
• Computer Vision: Multi-modal Vision

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation

Abstract

Constrained autoencoders (cAE) provide a successful path towards interpretable dimensionality reduction by enforcing geometric structure on latent spaces. However, standard cAEs cannot adapt to varying physical parameters or environmental conditions without conflating these contextual shifts with the primary input. To address this, we integrated a neuromodulatory mechanism into the cAE framework to allow for context-dependent manifold learning. This paper introduces the Neuromodulated Constrained Autoencoder (NcAE), which adaptively parameterizes geometric constraints via gain and bias tuning conditioned on static contextual information. Experimental results on dynamical systems show that the NcAE accurately captures how manifold geometry varies across different regimes while maintaining rigorous projection properties. These results demonstrate that neuromodulation effectively decouples global contextual parameters from local manifold representations. This architecture provides a foundation for developing more flexible, physics-informed representations in systems subject to (non-stationary) environmental constraints.

Problem

• context-dependent manifold learning
• constrained autoencoder
• neuromodulation
• non-stationary environments
• latent space geometry

Innovation

• neuromodulation
• constrained autoencoder
• context-dependent manifold learning
• geometric constraints
• non-stationary environments