Université du Travail Paul Pastur

An important class of neuro-symbolic (NeSy) methods relies on knowledge compilation (KC) techniques to transform logical constraints into a differentiable exact arithmetic circuit (AC) that represents all models of a logical formula. However, given the complexity of KC, compiling such exact circuits can be infeasible. Previous works in such cases proposed to compile a circuit for a subset of models. In this work, we will show that gradients calculated on a subset of models can be very far from true gradients. We propose a new framework that calculates gradients based on compiling logical constraints partially in not only a lower-bound circuit but also an upper-bound circuit. We prove that from this pair of ACs, gradients that are within a bounded distance from true gradients can be calculated. Our experiments show that adding the upper-bound AC also helps the learning process in practice, allowing for similar or better generalisation than working solely with fully compiled ACs, even with less than 150 seconds of partial compilation.

Reinforcement learning from visual observations is a challenging problem with many real-world applications. Existing algorithms mostly rely on a single observation from a well-designed fixed camera that requires human knowledge. Recent studies learn from different viewpoints with multiple fixed cameras, but this incurs high computation and storage costs and may not guarantee the coverage of the optimal viewpoint. To alleviate these limitations, we propose a straightforward View-conditional Partially Observable Markov Decision Processes (VPOMDPs) assumption and develop a new method, the MOdel-based SEnsor controlleR (MOSER). MOSER jointly learns a view-conditional world model (VWM) to simulate the environment, a sensory policy to control the camera, and a motor policy to complete tasks. We design intrinsic rewards from the VWM without additional modules to guide the sensory policy to adjust the camera parameters. Experiments on locomotion and manipulation tasks demonstrate that MOSER autonomously discovers task-specific viewpoints and significantly outperforms most baseline methods.

Spatial tuning of goal-directed movements is a critical phenomenon for the central nervous system. The use of reliable models is a necessary step for the understanding of the neural mechanisms governing reaching movements. We analysed phasic electromyographic (EMG) activities associated with upper limb reaching movements in the vertical plane (12 targets) in seven healthy subjects during three successive sessions.We tested the hypothesis that one of the fundamental parameters of the directional tuning of phasic EMG activities is the peak of EMG activity. To characterize the tuning of phasic EMG activities for the seven recorded muscles (brachioradialis, biceps, medial and long head of triceps, anterior and posterior deltoid, latissimus dorsi), we digitally compressed the EMG activities corresponding to slow reaches to the same targets into the time frame of the fast EMG traces. Estimates of gravity-related components were subtracted. Peak of EMG activities in the resulting phasic traces were identified for each muscle and each target. The maximal peak EMG amongst the 12 peaks of EMG activity in the sagittal plane was called M Peak EMG. We defined the directional dominance as the ratio of the M Peak EMG divided by the Peak EMG in the opposite direction. The M Peak EMG for each muscle was always found in the same privileged direction of the movement, except for the latissimus dorsi. The directional dominance remained unchanged across recording sessions for the brachioradialis, anterior deltoid and posterior deltoid. However, the directional dominance increased across sessions for the biceps, medial head of triceps, long head of triceps and latissimus dorsi. The invariability of the privileged direction of the M peak EMG was thus observed for six of the seven muscles investigated. Therefore, we suggest that the selection of the privileged direction of the M Peak EMG represents a robust parameter to study the neuromuscular control strategy underlying the specification of movement direction.