Low-power analogue neural networks with trainable nonlinear connections for continuous control

arXiv:2606.23742 · cs.LG, cs.AI, cs.AR · Submitted 2026-06-21 · Read on arXiv

cs.LG, cs.AI, cs.AR

Submitted: 2026-06-21

Updated: 2026-09-29

Comments: Preprint

Code: https://github.com/eleni-vasilaki/PhyKAN

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Terminology

• Fidelity — 1× in this paper · explained in 2 episodes

Sources

• Small-scale photonic Kolmogorov-Arnold networks using standard telecom nonlinear modules
• Learning Nonlinear Heterogeneity in Physical Kolmogorov-Arnold Networks
• Physical Analogue Kolmogorov-Arnold Networks based on Reconfigurable Nonlinear-Processing Units
• Estimating or Propagating Gradients Through Stochastic Neurons for Conditional Computation
• OpenAI Gym

Related papers

• Polynomial-Augmented Neural Networks (PANNs) with Weak Orthogonality Constraints for Enhanced Function and PDE Approximation
• AIRL-S: Unifying Reinforcement Learning and Search-Based Test-Time Scaling via Adversarial Inverse Reinforcement Learning
• Transformers as Bayesian In-Context Experimenters: Smoothness-Adaptive Efficient ATE Estimation
• Convergence issues in Relational Concept Analysis based on AOC-posets
• Beliefs Beyond Posteriors: The Predictive Cost of ELBO Binding
• Understanding Diffusion Models via Ratio-Based Function Approximation with SignReLU Networks