Joram Soch

This paper introduces multivariate Bayesian inversion as an alternative to machine learning for supervised tasks, demonstrating improved interpretability and probabilistic predictions.

This study uses 7 Tesla fMRI to show that 11-12-year-old children and adults have topographic brain responses tuned to numerical quantities, similar to adults.

The paper introduces and validates a method for fMRI decoding that accounts for trial-by-trial correlations using inverse transformed encoding models and searchlight analysis.

The paper analyzes fMRI data to show that visual working memory and sensory distractors are encoded in distinct but overlapping coding spaces, with dynamic changes over time.

The paper examines age-related differences in fMRI memory effects and links them to grey matter volume changes in older adults.

This study investigates how continuous perceptual choices are encoded in human early visual cortex using motion stimuli and encoding models.

The paper investigates effective connectivity patterns during explicit memory encoding in young and older adults, finding reduced inhibitory connectivity in older adults correlates with poorer memory performance.

The paper develops single-value fMRI scores to capture age-related differences in memory-related brain activity and links them to cognitive function in older adults.

The paper investigates whether working memory signals in early visual cortex depend on visual imagery, finding that reconstruction quality is similar for strong and weak imagers.

The paper examines how Openness to Experience relates to memory performance and neural activity in older adults, finding that higher Openness is linked to better memory and preserved brain network function.

The paper uses SVM to predict age and memory performance from structural and functional MRI data, finding gray matter volume best predicts age while task-based fMRI contrasts best predict memory performance.

The paper tests decoding behavior from brain signals by first reconstructing experimental conditions, finding this indirect approach works well despite being outperformed by direct decoding.

The paper introduces and validates a comprehensive fMRI score, FADE-SAME, to quantify memory-related neural activity changes associated with neurocognitive aging.

This study tests whether people can suppress their readiness potential (RP) through neurofeedback, finding no evidence of conscious control over RP magnitude during voluntary movements.

This study uses Bayesian model selection to show parametric models better explain fMRI encoding variance than categorical models in young and older adults.

This study shows that distributional transformation improves age prediction from MRI data when using deep regression, but not with linear regression in low-dimensional settings.

This paper introduces ITEM, a method to improve decoding accuracy in fMRI by addressing serial correlations in trial-wise estimates.

MACS provides a toolbox for assessing, comparing, and selecting GLMs in fMRI analysis, improving reproducibility and reducing false positives.

This paper introduces cross-validated Bayesian model averaging (cvBMA) to improve parameter estimates in GLM-based fMRI analysis by combining information from all models using posterior probabilities.