Training time logit averaging improves brain computer interface accuracy
Logit Averaging During Training Sets a New State of the Art for Non-Invasive Speech BCI
Machine Learning
Summary
Non-invasive brain-computer interfaces (BCIs) struggle to understand brain signals clearly because the signals are noisy. The authors compared different ways of averaging repeated brain signal trials during training and found that averaging the model’s predictions (logits) during training works best. This method sets a new accuracy record for recognizing speech sounds from brain signals. Their work highlights that how trials are combined is an important design choice for improving BCI performance.
What this means in practice
- •For neural interface developers: Improve speech decoding accuracy in brain-computer interfaces by using training-time logit averaging for multi-trial neural signal aggregation.
- •For speech recognition engineers: Enhance phoneme classification models by incorporating logit averaging methods to aggregate neural data from repeated trials during training.
Authors
Miran Özdogan, Gereon Elvers, Michael Bronstein, Oiwi Parker Jones
Abstract
Non-invasive brain-computer interfaces are in practice limited by signal-to-noise ratio, and recent state-of-the-art methods address this limitation by aggregating over repeated trials of the same stimulus. Yet multi-trial aggregation has been treated inconsistently across the literature: some work averages brain signals at the model's input, other work averages predictions at the model's output, and neither treats aggregation as an architectural choice in its own right. We provide the first systematic comparison of aggregation strategies for non-invasive neural decoding, evaluating six paradigms (including logit averaging during training and three attention-based variants we introduce) across three datasets, four tasks, and two group sizes. We find that aggregation choice matters substantially: training-time logit averaging, the simplest method we propose, outperforms all others in 6 of 7 settings. On the phoneme classification setting of the 2025 PNPL competition, it sets a new state of the art, significantly outperforming the competition winner by 17.4 absolute percentage points in balanced accuracy. We further draw a theoretical connection between logit averaging and joint ensemble training, which has been shown to underperform independent ensembling in the deep learning literature; our results call into question recent explanations of why joint ensemble training fails. Together, these findings establish multi-trial aggregation as a key design choice for non-invasive neural decoding and identify training-time logit averaging as a simple, effective default.