Training Spiking Neural Networks with Event-driven Backpropagation

Authors: Yaoyu Zhu, Zhaofei Yu, Wei Fang, Xiaodong Xie, Tiejun Huang, Timothée Masquelier

NeurIPS 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental The experimental results show that the proposed approach achieves state-of-the-art performance on CIFAR10 among time-based training methods. [...] In this section, we validate the effectiveness of our method on MNIST [48], Fashion-MNIST [49], N-MNIST [50], CIFAR10 [51], and CIFAR100 [51] datasets. This section is organized as follows: We first introduce the training details, then evaluate the performance of our algorithm and compare it with the state-of-the-art event-driven learning approaches. At last, we conduct ablation studies to illustrate the effectiveness of our proposed modules.
Researcher Affiliation Academia 1School of Computer Science, Peking University 2Institute for Artificial Intelligence, Peking University 3Centre de Recherche Cerveau et Cognition (CERCO), UMR5549 CNRS Univ. Toulouse 3, Toulouse, France
Pseudocode No The paper describes its learning rules and formulas in text and mathematical notation, but does not include any explicitly labeled 'Pseudocode' or 'Algorithm' blocks.
Open Source Code Yes Our code is available at https://github.com/zhuyaoyu/SNN-event-driven-learning.
Open Datasets Yes In this section, we validate the effectiveness of our method on MNIST [48], Fashion-MNIST [49], N-MNIST [50], CIFAR10 [51], and CIFAR100 [51] datasets. [...] The datasets we used in this paper are public.
Dataset Splits Yes Did you specify all the training details (e.g., data splits, hyperparameters, how they were chosen)? [Yes] [...] The test accuracy of these different settings is shown in Tab. 3.
Hardware Specification Yes We run all experiments on a single Nvidia A100 GPU.
Software Dependencies No The paper does not provide specific version numbers for software dependencies (e.g., Python, PyTorch, TensorFlow versions).
Experiment Setup Yes 4.1 Training Details. Initialization: When training in an event-driven fashion... Supervisory signals: Another problem we face is that output neurons corresponding to certain classes do not fire anymore... Experiment settings: In our experiments, we use the real-valued spike current representing the pixel intensities of the image as inputs. [...] Did you specify all the training details (e.g., data splits, hyperparameters, how they were chosen)? [Yes]