A mathematical model for the two-learners problem

Abstract

The results presented in this mathematical study allow the computation of simple theoretical simulations and performance of real experimental paradigms. Additionally, they are nicely in line with previous results in the BCI literature.

Brain-Machine Interfaces: A Tale of Two Learners

“…In conclusion, the promise of "zero-training" BMI and universal access to the technology for all prospective users remain elusive. There is thus mounting evidence pointing to the need of shifting the focus of investigation towards subject learning and the interactions between human and machine adaptation [12,13,16,32,33].…”

“…However, neural signals (and the features computed on those) are notorious for violating such assumptions. Non-stationarity effects have been well described [16,37,42]. Violation of the independence assumption, and a potentially varying degree of it over time, may invalidate classification performance estimation through techniques like crossvalidation and training-testing split [73], explaining potential "spurious" performance improvements that in fact do not represent subject learning.…”

“…Although the need for mutual learning is widely acknowledged in both invasive and non-invasive BMI [12][13][14][15][16], current research trends are heavily biased towards the machine learning side of BMI training. Here, we take a critical view of the relevant literature and our own previous work, in order to identify key issues for more effective mutual learning schemes in translational BMIs.…”

Regression Networks for Neurophysiological Indicator Evaluation in Practicing Motor Imagery Tasks

“…One more enhancing strategy of learning is identifying the causes of variability and incorporating appropriate actions in order to compensate for the BCI inefficiency [ 14 ], for instance, by including a calibration module that works hand-in-hand with the training procedure to make learning algorithms adapt to user EEG patterns [ 15 , 16 ]. In this regard, the correlation between the neural activity features that are extracted in advance (electrophysiological indicators or predictor) with the MI onset responses instructed via sensory stimuli can be assessed to prescreen participants for the ability to learn regulation of brain activity (pre-training measures) or for the improvement of learning abilities (training phase) [ 17 ].…”

A BCI challenge for the signal-processing community: considering the user in the loop

“…Indeed, both the user and the machine need to adapt to each other -the so-called co-adaptation in BCI [98]. A very recent and interesting work proposed a simple computational model to represent this interplay between the user learning and the machine learning, and how this co-adaptation takes place [99]. While such work is only a simulation, it has nonetheless suggested that an adaptation speed that is either too fast or too slow prevents this co-adaptation from converging, and leads to a decreased learning and performance.…”