Brokoslaw Laschowski

This paper introduces a framework for interpreting reward functions from inverse reinforcement learning, demonstrating its application to multi-agent social interactions.

This study evaluates diffusion latent representations for neural decoding, showing varying performance based on diffusion timesteps.

This paper introduces a brain-inspired framework for memory prioritization in neural networks based on valence, showing improved long-term memory and mitigation of the 'lost-in-the-middle' problem in language modeling.

This paper develops an EMG foundation model for neural decoding that outperforms subject-optimized models in both in-distribution and out-of-distribution evaluations.

This paper introduces a mixture model framework for domain-adaptive brain decoding, achieving state-of-the-art performance with less training data by optimizing mixture weights via convex optimization.

This paper develops a real-time neural controller using an EMG-driven musculoskeletal model to enable volitional control of robots and computers with high accuracy and low latency.

This paper develops a multimodal vision-language system for egocentric perception of walking environments, generating detailed image captions with user personalization.

The paper compares neural decoding algorithms for brain-machine interfaces to find the optimal combination of signal processing, feature extraction, and classifiers.

The paper compares static and sequential image classification of walking environments using temporal neural networks, finding that 3D models outperform 2D ones but have higher computational demands.

This paper develops a two-layer Q-learning algorithm for simulating human locomotion, achieving 79% accuracy and improved efficiency compared to previous methods.

This paper develops AI-powered smart glasses for real-time environment sensing and recognition in human-robot walking scenarios, achieving 93.6% accuracy with low latency.