Real-time reinforcement for human-machine interface control - UPHUMMEL - EPFL

Dataset and code to reproduce the main results of the paper Real-time reinforcement for human-machine interface control, Neuron (2026). by Pierre Vassiliadis, Daniel Leal Pinheiro, Lisa Fleury, Alexandre Zenon, Martín Esparza-Iaizzo, Abigaïl Ingster, Silvestro Micera, Solaiman Shokur, Friedhelm C Hummel. This repository contains MATLAB code and block-wise data organization used for the analyses of Experiment 1. Repository content Data are organized by participant, with one folder per subject. Each subject folder contains main task files of the form: rml_FTT_B0_*.mat to rml_FTT_B8_*.mat as well as titration files: rml_FTT_titration_B1_*.matrml_FTT_titration_B2_*.matrml_FTT_titration_B3_*.mat For the main task, blocks B0, B1, and B2 correspond to different familiarisation blocks (1: sinusoid sequence, 2: different sequence, 0: short familiarisation with the visual uncertainty and with or without reinforcement), whereas blocks B3–B8 correspond to the main experimental blocks. Trials are divided into four phases: sinusoidpre-trainingtrainingpost-training Condition coding in the analysis script is: 1 = Reinforcement OFF, full vision2 = Reinforcement OFF, high vision3 = Reinforcement OFF, low vision4 = Reinforcement ON, full vision5 = Reinforcement ON, high vision6 = Reinforcement ON, low vision The provided MATLAB script reproduces the main analysis pipeline for Experiment 1, including trial-level table construction, frame-level variable extraction, learning metrics, acceleration-based analyses, and figure generation. The rml_FTT_titration_*.mat files can be used to analyze Experiment 2. The code was cleaned for public sharing while preserving the original analysis logic. Some external MATLAB helper functions may be required for specific sections of the script, including shadedErrorBar, violin, diff2pt, and computeKL. Contact: [e-mail], [e-mail]