Topic: logistic regression

Stringer et al. used two-photon calcium imaging to record activity from tens of thousands of V1 cells in head-fixed mice during a visual orientation discrimination task, and showed that the neural response encodes stimulus discrimination thresholds nearly 100 times more precise than the corresponding behavioral thresholds. Animal arousal state was monitored by measuring the locomotion speed during task execution, but could not explain this large discrepancy in discrimination threshold. Here, we take advantage of the fact that the behavioral data was acquired in that experiment, and ask whethe

Generalized linear models (GLMs) are a gold-standard in computational neuroscience, allowing for the statistical investigation of neural activity when a population of neurons is exposed to stimuli (Gerwinn, 2010 et al.). Logistic regression models are one of such GLMs that allow for the prediction of two possible outcomes (i.e. binary dependent variables) based on predictor variables (i.e. the neural activity). One of such neural activity of interest is the one produced during language-related tasks in a study done by the Human Connectome Project (Binder et. al, 2020). In this study, the subje

While non-invasive Brain-to-Computer interface can accurately classify the lateralization of hand moments, the distinction of fingers activation in the same hand is limited by their local and overlapping representation in the motor cortex. In particular, the low signal-to-noise ratio restrains the opportunity to identify meaningful patterns in a supervised fashion. Here we combined Magnetoencephalography (MEG) recordings with advanced decoding strategy to classify finger movements at single trial level. We recorded eight subjects performing a serial reaction time task, where they pressed four