Direct Training of SNN using Local Zeroth Order Method

Authors: Bhaskar Mukhoty, Velibor Bojkovic, William de Vazelhes, Xiaohan Zhao, Giulia De Masi, Huan Xiong, Bin Gu

NeurIPS 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We perform experimental validation of the technique on standard static datasets (CIFAR-10, CIFAR-100, Image Net100) and neuromorphic datasets (DVS-CIFAR-10, DVS-Gesture, N-Caltech-101, NCARS) and obtain results that offer improvement over the state-of-the-art results.
Researcher Affiliation Collaboration 1 Mohamed bin Zayed University of Artificial Intelligence, UAE 2 ARRC, Technology Innovation Institute, UAE 3 Nanjing University of Information Science and Technology, China 4 School of Artificial Intelligence, Jilin University, China 5 Harbin Institute of Technology, China 6 Bio Robotics Institute, Sant Anna School of Advanced Studies, Pisa, Italy
Pseudocode Yes Algorithm 1 LOCALZO
Open Source Code Yes The code is available at https://github.com/Bhaskar Mukhoty/Local ZO.
Open Datasets Yes standard static image datasets such as CIFAR-10, CIFAR-100[22], Image Net-100[12] and neuromorphic datasets such as DVS-CIFAR-10[24], DVS-Gesture[2], N-Caltech-101[30], N-CARS[37].
Dataset Splits No The paper provides train and test image counts for datasets like CIFAR-10, CIFAR-100, and ImageNet-100 (e.g., 'each class respectively have (5000, 1000) train and test images' for CIFAR-10), but does not explicitly detail the size or methodology for a validation set split.
Hardware Specification No The paper does not provide specific hardware details (such as GPU or CPU models, or memory specifications) used to run its experiments.
Software Dependencies No The paper mentions 'Optimizer: Adam' but does not specify version numbers for any programming languages, libraries, or other software dependencies.
Experiment Setup Yes Table 7 provides 'Hyper-parameter settings for general comparison' including 'Number epochs', 'Mini batch size', 'Learning Rate', 'Optimizer: Adam with betas: (0.9; 0.999)', and specific parameters for LIF and LOCALZO (δ, m, λ).