This dataset is based on the Spiking Heidelberg Digits (SHD) dataset. Sample inputs consist of two spike encoded digits sampled uniformly at random from the SHD dataset and concatenated, with the target being the sum of the digits (irrespective of language). The train and test split remain the same, with the test set consisting of 16k such samples based on the SHD test set.

For comparability, please report the performance for a temporal binning resolution of 2ms, and use a last time step loss to test the model’s temporal integration capabilities.

Importantly, solving this dataset requires integrating temporal information over multiple timescales; on a shorter timescale identifying each digit, on a longer timescale calculating their sum, crucially requiring retaining the first digit in memory.

When using this dataset, please cite the following two papers:

[1] Spieler, A., Rahaman, N., Martius, G., Schölkopf, B., & Levina, A. (2023). The ELM Neuron: an Efficient and Expressive Cortical Neuron Model Can Solve Long-Horizon Tasks. arXiv preprint arXiv:2306.16922.

[2] Cramer, B., Stradmann, Y., Schemmel, J., & Zenke, F. (2020). The heidelberg spiking data sets for the systematic evaluation of spiking neural networks. IEEE Transactions on Neural Networks and Learning Systems, 33(7), 2744-2757.