Search papers, labs, and topics across Lattice.
The paper introduces Collaborative Temporal Feature Generation (CTFG), a reinforcement learning framework for cross-user human activity recognition using wearable sensor data. CTFG uses a Transformer-based autoregressive generator to create feature token sequences optimized via Group-Relative Policy Optimization, a novel critic-free RL algorithm that evaluates generated sequences against alternatives from the same input. Experiments on DSADS and PAMAP2 datasets demonstrate state-of-the-art cross-user accuracy, reduced training variance, and accelerated convergence compared to existing domain generalization methods.
Ditch the critic: This new reinforcement learning approach trains feature extractors for human activity recognition without needing a value function, leading to more stable and generalizable performance across diverse users.
Human Activity Recognition using wearable inertial sensors is foundational to healthcare monitoring, fitness analytics, and context-aware computing, yet its deployment is hindered by cross-user variability arising from heterogeneous physiological traits, motor habits, and sensor placements. Existing domain generalization approaches either neglect temporal dependencies in sensor streams or depend on impractical target-domain annotations. We propose a different paradigm: modeling generalizable feature extraction as a collaborative sequential generation process governed by reinforcement learning. Our framework, CTFG (Collaborative Temporal Feature Generation), employs a Transformer-based autoregressive generator that incrementally constructs feature token sequences, each conditioned on prior context and the encoded sensor input. The generator is optimized via Group-Relative Policy Optimization, a critic-free algorithm that evaluates each generated sequence against a cohort of alternatives sampled from the same input, deriving advantages through intra-group normalization rather than learned value estimation. This design eliminates the distribution-dependent bias inherent in critic-based methods and provides self-calibrating optimization signals that remain stable across heterogeneous user distributions. A tri-objective reward comprising class discrimination, cross-user invariance, and temporal fidelity jointly shapes the feature space to separate activities, align user distributions, and preserve fine-grained temporal content. Evaluations on the DSADS and PAMAP2 benchmarks demonstrate state-of-the-art cross-user accuracy (88.53\% and 75.22\%), substantial reduction in inter-task training variance, accelerated convergence, and robust generalization under varying action-space dimensionalities.
