Category: [A1]; Team name: MaAnCla; Dataset: Dynamical Activity Complex (DAC)

[andrea-cavallo-98]

Checklist

• My pull request has a clear and explanatory title.
• My pull request passes the Linting test.
• I added appropriate unit tests and I made sure the code passes all unit tests. (refer to comment below)
• My PR follows PEP8 guidelines. (refer to comment below)
• My code is properly documented, using numpy docs conventions, and I made sure the documentation renders properly.
• I linked to issues and PRs that are relevant to this PR.

Description

This pull request adds support to the Dynamical Activity Complex (DAC) datasets from the paper [1]. These datasets contain structural brain directed graphs and neuronal activations in response to different stimuli.
Specifically, the PR implements a dataset and a dataloader class, considers 3 different datasets (named 4-325-1, 4-325-3, 4-325-5, corresponding, respectively, to the datasets M=1, M=3, M=5 in [1]) and loads the predefined 5 splits from [1] (which can be selected via the configuration parameter split_num).

Dataset description

The DAC digraphs are derived from the Blue Brain Project's reconstruction of a microcircuit in the somatosensory cortex of a rat [2]. For a specific graph, each node corresponds to a neuron, directed edges correspond to neuronal connections and node features (of shape $T$) contain the activation state of a neuron (0 or 1) over $T$ time bins. Activations are obtained by injecting one of eight distinct synaptic input patterns delivered via thalamocortical afferents, and the dataset task consists in classifying which of the eight stimula the observed activations are associated to. To generate different graphs, subgraphs of different size and the associated activations are subsampled, resulting in the 3 datasets of this PR.

Issue

Neuronal activations are naturally modeled by networks and generate higher-order (directed) patterns informative of brain dynamics, making these datasets of high interest for the TDL community.

References

[1] Lecha, Manuel, et al. "Directed Semi-Simplicial Learning with Applications to Brain Activity Decoding." arXiv preprint arXiv:2505.17939 (2025).

[2] Markram, Henry, et al. "Reconstruction and simulation of neocortical microcircuitry." Cell 163.2 (2015): 456-492.