Learning from label proportions in brain-computer interfaces: Online unsupervised learning with guarantees

Abstract

ObjectiveUsing traditional approaches, a brain-computer interface (BCI) requires the collection of calibration data for new subjects prior to online use. Calibration time can be reduced or eliminated e.g., by subject-to-subject transfer of a pre-trained classifier or unsupervised adaptive classification methods which learn from scratch and adapt over time. While such heuristics work well in practice, none of them can provide theoretical guarantees. Our objective is to modify an event-related potential (ERP) paradigm to work in unison with the machine learning decoder, and thus to achieve a reliable unsupervised calibrationless decoding with a guarantee to recover the true class means.MethodWe introduce learning from label proportions (LLP) to the BCI community as a new unsupervised, and easy-to-implement classification approach for ERP-based BCIs. The LLP estimates the mean target and non-target responses based on known proportions of these two classes in different groups of the data. We present a visual ERP speller to meet the requirements of LLP. For evaluation, we ran simulations on artificially created data sets and conducted an online BCI study with 13 subjects performing a copy-spelling task.ResultsTheoretical considerations show that LLP is guaranteed to minimize the loss function similar to a corresponding supervised classifier. LLP performed well in simulations and in the online application, where 84.5% of characters were spelled correctly on average without prior calibration.SignificanceThe continuously adapting LLP classifier is the first unsupervised decoder for ERP BCIs guaranteed to find the optimal decoder. This makes it an ideal solution to avoid tedious calibration sessions. Additionally, LLP works on complementary principles compared to existing unsupervised methods, opening the door for their further enhancement when combined with LLP.

From full calibration to zero training for a code-modulated visual evoked potentials brain computer interface

“…Despite theoretical advances, practical applications of transfer learning remain scarce, especially for evoked-response BCIs. For the P300 speller, several attempts have been made using unsupervised learning and online adaptation [23,24] and learning from label proportions [25]. An explicit attempt to reduce the training time for cVEP BCIs was made using an automatic repeat request, which stops acquiring more training trials when a certain reliability measurement reaches a threshold during calibration [26].…”

The combination of brain-computer interfaces and artificial intelligence: applications and challenges

“…The use of implantable electrodes allows patients to control movements with several degrees of freedom, enabling them to make more complex and functional movements. However, approaches that use noninvasive systems provide limited control, and most complex movements rely on the AI of the robot (40). To classify motor-related signals specifically for BCI applications, Nurse and colleagues developed a generalized approach that takes advantage of a stochastic machine-learning method (41).…”

A Robust Screen-Free Brain-Computer Interface for Robotic Object Selection

“…While multiple signal categories such as mental imagery can provide information, here we focus on experimental paradigms utilizing event-related potentials in response to a presented stimulus. Examples include spelling using visual (Sellers and Donchin, 2006 ; Hübner et al, 2017 ; Nagel and Spüler, 2019 ), auditory (Schreuder et al, 2010 ) or tactile (van der Waal et al, 2012 ) information and control of devices (Tangermann et al, 2009 ). In order to classify individual ERP responses, mean electrode potentials in suitable time intervals after the stimulus can be combined with linear classification methods (Blankertz et al, 2011 ).…”