Progressive Feature Interaction Search for Deep Sparse Network

Authors: Chen Gao, Yinfeng Li, Quanming Yao, Depeng Jin, Yong Li

NeurIPS 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experiments on three real-world benchmark datasets show promising results of PROFIT in both accuracy and efficiency.
Researcher Affiliation Collaboration 1Beijing National Research Center for Information Science and Technology, Department of Electronic Engineering, Tsinghua University 24Paradigm Inc.
Pseudocode Yes Algorithm 1 Progressive gradient descent.
Open Source Code No The paper provides links to baseline implementations but does not provide concrete access to its own source code, nor does it explicitly state that its own code will be released.
Open Datasets Yes To validate the effectiveness of our proposed PROFIT, we conduct experiments on three benchmark datasets (Criteo, Avazu and ML1M) widely used in existing works of deep sparse networks [5, 31, 18] to evaluate the performance, of which the details are provided in the Appendix.
Dataset Splits Yes where Dtra and Dval denote the training and validation datasets, respectively. All the other hyper-parameters are tuned on the validation set.
Hardware Specification No The paper mentions 'our normal hardware platform' but does not provide specific details such as GPU models, CPU types, or memory amounts used for running experiments.
Software Dependencies No The paper states 'We implement our methods using Py Torch' but does not specify the version number for PyTorch or any other software dependencies.
Experiment Setup Yes We apply Adam with a learning rate of 0.001 and a mini-batch size of 4096, a widely-used setting in existing works [5, 31]. We set the embedding sizes to 16 in all the models. We use the same neural network structure ({400, 400, 400}) for all methods that adopt MLP for a fair comparison, following [5, 31].