Renate Krause

This study shows that low initial weight variance in recurrent neural networks leads to overlapping weight subspaces for subtasks, while high variance results in more segregated subspaces.

This study uses data-driven methods to reduce the dimensionality of RNNs and identify core dynamics, leading to new insights into neuronal decision-making and memory.

This paper identifies low-dimensional 'operative dimensions' in RNNs that significantly influence dynamics, showing that much of the high-dimensional connectivity is functionally irrelevant.

The paper presents an automatic method to identify synaptic partners in a whole-brain Drosophila EM dataset using convolutional neural networks, achieving high accuracy in predicting synaptic connections.