Dataset Information

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

ABSTRACT:

Objective

Using 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.

Method

We 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.

Results

Theoretical 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.

Significance

The 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.

SUBMITTER: Hubner D

PROVIDER: S-EPMC5391120 | biostudies-literature | 2017

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

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

Hübner David D Verhoeven Thibault T Schmid Konstantin K Müller Klaus-Robert KR Tangermann Michael M Kindermans Pieter-Jan PJ

PloS one 20170413 4

<h4>Objective</h4>Using 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-re ...[more]

PMID: 28407016

Similar Datasets

Project description:BACKGROUND:Individuals with severe neurological disabilities but preserved cognition, including children, are often precluded from connecting with their environments. Brain computer interfaces (BCI) are a potential solution where advancing technologies create new clinical opportunities. We evaluated clinician awareness as a modifiable barrier to progress and identified eligible populations. METHODS:We executed a national, population-based, cross-sectional survey of physician specialists caring for persons with severe disability. An evidence- and experience-based survey had three themes: clinician BCI knowledge, eligible populations, and potential impact. A BCI knowledge index was created and scored. Canadian adult and pediatric neurologists, physiatrists and a subset of developmental pediatricians were contacted. Secure, web-based software administered the survey via email with online data collection. RESULTS:Of 922 valid emails (664 neurologists, 253 physiatrists), 137 (15%) responded. One third estimated that ≥10% of their patients had severe neurological disability with cognitive capacity. BCI knowledge scores were low with > 40% identifying as less than "vaguely aware" and only 15% as "somewhat familiar" or better. Knowledge did not differ across specialties. Only 6 physicians (4%) had patients using BCI. Communication and wheelchair control rated highest for potentially improving quality of life. Most (81%) felt BCI had high potential to improve quality of life. Estimates suggested that > 13,000 Canadians (36 M population) might benefit from BCI technologies. CONCLUSIONS:Despite high potential and thousands of patients who might benefit, BCI awareness among clinicians caring for disabled persons is poor. Further, functional priorities for BCI applications may differ between medical professionals and potential BCI users, perhaps reflecting that clinicians possess a less accurate understanding of the desires and needs of potential end-users. Improving knowledge and engaging both clinicians and patients could facilitate BCI program development to improve patient outcomes.

Dataset Information

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

Objective

Method

Results

Significance

Publications

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

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets