Adaptive Boosting of DNN Ensembles for Brain-Computer Interface Spellers

Güney, Osman Berke; Koc, Emirhan; Aksoy, Ömer Can; Catak, Yigit; Arslan, Suayb S.; Özkan, Hüseyin

doi:10.1109/siu53274.2021.9477841

Cited by 1 publication

(3 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Although the methods in the first group (e.g. [11,13,16]) typically yield better target identification accuracy, they (unlike ours) require lengthy and tiring user-specific training. Hence, this first group of methods are not comparable to our proposed method here.…”

Section: Related Workmentioning

confidence: 97%

“…In contrast, in this presented study, we combined only k most representative fine-tuned DNNs based on a weighting devised through a novel correlation based similarity measure and dynamic selection, which is fundamentally different. Moreover, [16] requires user-specific training thus belongs to the first group, whereas the presented study does not.…”

Section: Related Workmentioning

confidence: 99%

“…Hence, this first group of methods are not comparable to our proposed method here. In our preliminary short conference proceeding (in Turkish) [16], we used the same ensemble of fined-tuned DNNs but combined all of them with weights given by a variant of the AdaBoost algorithm. In contrast, in this presented study, we combined only k most representative fine-tuned DNNs based on a weighting devised through a novel correlation based similarity measure and dynamic selection, which is fundamentally different.…”

Section: Related Workmentioning

confidence: 99%

See 2 more Smart Citations

Transfer learning of an ensemble of DNNs for SSVEP BCI spellers without user-specific training

Güney¹,

Özkan²

2023

J. Neural Eng.

View full text Add to dashboard Cite

Objective: Steady-state visually evoked potentials (SSVEPs), measured with EEG (electroencephalogram), yield decent information transfer rates (ITR) in brain-computer interface (BCI) spellers. However, the current high performing SSVEP BCI spellers in the literature require an initial lengthy and tiring user-speciﬁc training for each new user for system adaptation, including data collection with EEG experiments, algorithm training and calibration (all are before the actual use of the system). This impedes the widespread use of BCIs. To ensure practicality, we propose a highly novel target identiﬁcation method based on an ensemble of deep neural networks (DNNs), which does not require any sort of user-speciﬁc training. Approach: We exploit already-existing literature datasets from participants of previously conducted EEG experiments to train a global target identiﬁer DNN ﬁrst, which is then ﬁne-tuned to each participant. We transfer this ensemble of ﬁne-tuned DNNs to the new user instance, determine the k most representative DNNs according to the participants’ statistical similarities to the new user, and predict the target character through a weighted combination of the ensemble predictions. Main results: The proposed method signiﬁcantly outperforms all the state-of-the-art alternatives for all stimulation durations in [0.2 − 1.0] seconds on two large-scale benchmark and BETA datasets, and achieves impressive 155.51 bits/min and 114.64 bits/min ITRs. Code is available for reproducibility: https://github.com/osmanberke/Ensemble-of-DNNs Signiﬁcance: Our Ensemble-DNN method has the potential to promote the practical widespread deployment of BCI spellers in daily lives as we provide the highest performance while enabling the immediate system use without any user-speciﬁc training.

show abstract

Section: Related Workmentioning

confidence: 97%

Section: Related Workmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation