Training Neural Networks Using Features Replay

Authors: Zhouyuan Huo, Bin Gu, Heng Huang

NeurIPS 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Finally, we apply our method to training deep convolutional neural networks, and the experimental results show that the proposed method achieves faster convergence, lower memory consumption, and better generalization error than compared methods.
Researcher Affiliation Collaboration Zhouyuan Huo1,2, Bin Gu2, Heng Huang1,2 1Electrical and Computer Engineering, University of Pittsburgh, 2 JDDGlobal.com zhouyuan.huo@pitt.edu, jsgubin@gmail.com heng.huang@pitt.edu
Pseudocode Yes Algorithm 1 Features Replay Algorithm
Open Source Code No The paper does not provide an explicit statement or link for open-source code related to the described methodology.
Open Datasets Yes We implement our method in Py Torch [28], and evaluate it with Res Net models [8] on two image classification benchmark datasets: CIFAR-10 and CIFAR-100 [18].
Dataset Splits No The paper mentions using CIFAR-10 and CIFAR-100 datasets but does not explicitly provide the specific training, validation, and test dataset splits used for reproduction, nor does it cite a source for these specific splits.
Hardware Specification Yes All experiments are performed on a server with four Titan X GPUs.
Software Dependencies No The paper states 'We implement our method in Py Torch [28]', but does not provide specific version numbers for PyTorch or any other software dependencies.
Experiment Setup Yes We use SGD with the momentum of 0.9, and the stepsize is initialized to 0.01. Each model is trained using batch size 128 for 300 epochs and the stepsize is divided by a factor of 10 at 150 and 225 epochs. The weight decay constant is set to 5e-4.