Online Convolutional Sparse Coding with Sample-Dependent Dictionary
Authors: Yaqing Wang, Quanming Yao, James Tin-Yau Kwok, Lionel M. NI
ICML 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experimental results on a number of data sets show that the proposed method outperforms existing CSC algorithms with significantly reduced time and space complexities. |
| Researcher Affiliation | Collaboration | 1Department of Computer Science and Engineering, Hong Kong University of Science and Technology University, Hong Kong 24Paradigm Inc, Beijing, China. 3Department of Computer and Information Science, University of Macau, Macau. |
| Pseudocode | Yes | Algorithm 1 Sample-dependent CSC (SCSC). 1: Initialize W0 2 W, B0 2 B, H0 = 0, G0 = 0; 2: for t = 1, 2, . . . , T do 3: draw xt from {xi}; 4: xt = F(xt); 5: obtain Wt, Zt using ni APG; 6: for r = 1, 2, . . . , R do 7: Yt(:, r) = F(Zt W > t (:, r)); 8: end for 9: update { Ht(:, :, 1), . . . , Ht(:, :, P)} using (14); 10: update { Gt(:, 1), . . . , Gt(:, P)} using (15); 11: update Bt by (13) using ADMM; 12: end for 13: for r = 1, 2, . . . , R do 14: BT (:, r) = C(F 1( BT (:, r))); 15: end for output BT . |
| Open Source Code | No | The paper does not provide concrete access to source code for the methodology described. |
| Open Datasets | Yes | Experiments are performed on a number of data sets (Table 2). Fruit and City are two small image data sets that have been commonly used in the CSC literature (Zeiler et al., 2010; Bristow et al., 2013; Heide et al., 2015; Papyan et al., 2017). ...In some experiments, we will also use two larger data sets, CIFAR10 (Krizhevsky & Hinton, 2009) and Flower (Nilsback & Zisserman, 2008). |
| Dataset Splits | Yes | We use the default training and testing splits provided in (Bristow et al., 2013). ... Table 2. Summary of the image data sets used. size #training #testing Fruit 100 100 10 4 City 100 100 10 4 CIFAR-10 32 32 50,000 10,000 Flower 500 500 2,040 6,149 |
| Hardware Specification | Yes | Because of the small memory footprint of SCSC, we run it on a GTX 1080 Ti GPU in this experiment. |
| Software Dependencies | No | The paper does not provide specific software names with version numbers needed to replicate the experiment. |
| Experiment Setup | Yes | Following (Heide et al., 2015; Choudhury et al., 2017; Papyan et al., 2017; Wang et al., 2018), we set the filter size M as 11 11, and the regularization parameter β in (1) as 1. |