Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in [1].
Stochastic Regularized Majorization-Minimization with weakly convex and multi-convex surrogates
Authors: Hanbaek Lyu
JMLR 2024 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | In Section 5.2, we provide numerical experiments of SRMM on network dictionary learning and image classification using deep convolutional neural networks. |
| Researcher Affiliation | Academia | Hanbaek Lyu EMAIL Department of Mathematics University of Wisconsin Madison, WI 53706, USA |
| Pseudocode | Yes | In Section 3, we state our SRMM algorithm in Algorithm 1. Next in Section 4, we state our main results (Theorems 2, 3, and 4) and their corollaries (Corollaries 5 and 6). Algorithm 1 Stochastic Regularized Majorization-Minimization (SRMM) Algorithm 2 Surrogate Minimization with Proximal Regularization Algorithm 3 Block Surrogate Minimization with Diminishing Radius |
| Open Source Code | Yes | All codes are available at https://github.com/Hanbaek Lyu/SRMM |
| Open Datasets | Yes | and image classification with deep convolutional neural networks for the CIFAR-10 dataset Krizhevsky et al. (2009)... We consider three facebook networks of schools Caltech, UCLA, and Wisconsin from the Facebook100 dataset (Traud et al., 2012) |
| Dataset Splits | No | In this section, we provide experimental results of SRMM on two tasks Network Dictionary Learning ? and Image classification with Deep Convolutional Neural Networks for the CIFAR-10 dataset Krizhevsky et al. (2009). The paper mentions using the "standard CIFAR-10 dataset" but does not explicitly state specific dataset splits (e.g., percentages, counts, or explicit split files) for reproduction. |
| Hardware Specification | No | The paper does not provide specific hardware details (e.g., GPU/CPU models, memory) used for running its experiments. It discusses tasks and algorithms but not the underlying computing resources. |
| Software Dependencies | No | The paper does not provide specific software dependency details (e.g., library or solver names with version numbers like Python 3.8, PyTorch 1.9) needed to replicate the experiment. |
| Experiment Setup | Yes | In the experiments reported in Figure 5, the hyperparameters are chosen as follows. (1) SRMM: wn = n 1/2 for n 1, L {10, 1}, and λ {10, 1, 10 1, 10 2, 10 3}; (2) SMM: wn = n 1/2 for n 1 and L {10, 1}; (3) SGD-HB: step sizes in {10 2, 10 1, 1, 10, 102} and momentum parameter of 0.9; (4) Ada Grad: stepsizes in {5e-4, 1e-3, 5e-3, 5e-2, 1e-1} and accumulator value 0; (5)-(6) Adam and AMSGrad: step sizes in {1e-4, 5e-4, 1e-3, 5e-3, 1e-2}, β1 {0.9, 0.99}, and β2 {0.99, 0.999}. For SRMM and SMM, we reset the iteration counter to zero at the beginning of each epoch in order to avoid the adaptivity weight wn becoming too small after many epochs. |