IPAB Workshop - 25/6/26

Speaker: Florent Le Moel

Title: A Modular, Biophysically-Grounded Rendering Framework for Insect Vision Neuroethology

Abstract: Insects solve complex navigational tasks with remarkable efficiency, using minimal neural hardware
tuned to the specific requirements of their ecological niches. To truly understand or replicate these behaviours,
it is insufficient to model the brain in isolation: one must account for the dynamic, closed-loop
interactions between the environment, the physical organisation of the sensory periphery, and internal
biophysical dynamics. We present a modular, hardware-agnostic and high-performance
rendering framework specifically designed for insect neuroethology and neuromorphic research. Unlike
existing compound eye simulations that rely on proprietary vendor-locked architectures, ours
leverages a Python-centric philosophy with real-time ray-tracing and stochastic path-tracing on any
GPU architecture. Crucially, the engine moves beyond the static ’ommatidium-as-a-pixel’ paradigm
by introducing a fully parametrisable model where every layer of the compound eye (from the geo-
metric shape and the topological lattice to the internal rhabdomere blueprint) is a discrete, swappable
component. The engine is capable of simulating the high-frequency, sub-ommatidial rhabdomere pho-
tomechanical actuation, allowing for the investigation of a variety of active sensing phenomena within a
real-time closed-loop environment. The framework also includes an automated morphological pipeline
that allows transforming 2D anatomical data into faithful 3D sensory models. We validate the engine
through two case studies: a closed-loop optic-flow centring response in a virtual tunnel, and the recov-
ery of spatial hyperacuity via rhabdomere microsaccades. By providing a bridge between high-fidelity
visual ecology and neuromorphic modelling, this project enables researchers to explore how the in-
terplay of sensory optics and neural processing can generate complex behaviour in both biological and
artificial agents.

Speaker: Manisha Dubey

Title: Towards Human-Centered AI: From Temporal Dynamics to Preferences and Cognition 

Abstract: My research develops probabilistic and human-centered AI methods for learning from complex behavioural data and supporting decision-making under uncertainty. In this talk, I will present my research journey from modelling temporal event dynamics during my PhD, through preference-aware multi-objective Bayesian optimization and its application in sustainable process design, to my current work on developing environments to model human cognition and behaviour towards the broader goal of assistive autonomy. 

I will begin with my doctoral work on Hawkes process and neural temporal point processes, where the goal was to infer latent dynamics from event streams. I will then discuss my postdoctoral work on human-in-the-loop multi-objective Bayesian optimization, where AI systems learn the preferences of the decision-maker to navigate trade-offs, including sustainability-driven polymer process optimization. Finally, I will present my current work on modelling human behaviour using Bayesian experimental design, inverse planning and CogniCart - a virtual supermarket for modelling executive function and studying cognitive support. Viewed collectively, these projects represent a progression from modelling dynamic behaviour to understanding human objectives and cognitive processes, with the broader goal of enabling adaptive human-centred AI systems and their application to various applications like sustainability, healthcare, assistive autonomy and scientific discovery.