iiinova — where your digital self keeps evolving

marcelvangerven.bsky.social

@marcelvangerven.bsky.social

marcelvangerven.bsky.social @marcelvangerven.bsky.social · 08/05/2026

Proud to share a new pre-print by @davidleeftink.bsky.social and with @maxhinne.bsky.social. We make a formal connection between the Pontryagin minimum principle and deep recurrent reinforcement learning, improving performance on challenging control tasks! arxiv.org/pdf/2605.05373

Recurrent Reinforcement Learning via Neural co-state policies. NCP s are regularized during training to mirror the optimality conditions implied by the minimum principle. By structuring the hidden states as representations of the underlying optimal control co-states, the read-out layer acts as a control-Hamiltonian minimizer.

041

Reposted by @marcelvangerven.bsky.social

Max Hinne @maxhinne.bsky.social · 01/12/2025

Proud to share a new pre-print by @davidleeftink.bsky.social and with @marcelvangerven.bsky.social on Bayesian optimization for automatic laser dicing in semiconductor manufacturing! arxiv.org/abs/2511.23141

012

marcelvangerven.bsky.social @marcelvangerven.bsky.social · 19/09/2025

New preprint out on neuromorphic intelligence: arxiv.org/pdf/2509.11940

arxiv.org

000

Reposted by @marcelvangerven.bsky.social

David Leeftink @davidleeftink.bsky.social · 03/09/2025

Our recent paper “Optimal Control of Probabilistic Dynamics Models via Mean Hamiltonian Minimization,” has been accepted to CDC 2025! We demonstrate how applying optimal control principles can significantly improve planning in deep model-based reinforcement learning with epistemic uncertainty.

181

marcelvangerven.bsky.social @marcelvangerven.bsky.social · 05/04/2025

New preprint out! arxiv.org/abs/2504.02543

arxiv.org

Probabilistic Pontryagin's Maximum Principle for Continuous-Time Model-Based Reinforcement Learning

Without exact knowledge of the true system dynamics, optimal control of non-linear continuous-time systems requires careful treatment of epistemic uncertainty. In this work, we propose a probabilistic...

030