Learning the Positions in CountSketch

Authors: Yi Li, Honghao Lin, Simin Liu, Ali Vakilian, David Woodruff

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

Reproducibility Variable Result LLM Response
Research Type Experimental 7 EXPERIMENTS: LOW-RANK APPROXIMATION, 8 EXPERIMENTS: SECOND-ORDER OPTIMIZATION. Our empirical results are provided in Table 7.1 for both Algorithm 2 and Algorithm 1, where the errors take an average over 10 trials. We plot in Figures 7.1 the mean errors on a logarithmic scale.
Researcher Affiliation Academia Yi Li Nanyang Technological University yili@ntu.edu.sg Honghao Lin, Simin Liu Carnegie Mellon University {honghaol, siminliu}@andrew.cmu.edu Ali Vakilian Toyota Technological Institute at Chicago vakilian@ttic.edu David P. Woodruff Carnegie Mellon University dwoodruf@andrew.cmu.edu
Pseudocode Yes Algorithm 1 Rank-k approximation of A using a sketch S, Algorithm 2 ALGLRA(SKETCH-LOWRANK), Algorithm 3 POSITION OPTIMIZATION: GREEDY SEARCH, Algorithm 4 Position optimization: Inner Product, Algorithm 5 LRA APPROXCHECK
Open Source Code No The paper does not contain any explicit statements or links indicating that the source code for the described methodology is publicly available.
Open Datasets Yes We use the three datasets from (Indyk et al., 2019): (1, 2) Friends, Logo (image): frames from a short video of the TV show Friends and of a logo being painted; (3) Hyper (image): hyperspectral images from natural scenes. ... We use the Electric1 dataset of residential electric load measurements. ... https://archive.ics.uci.edu/ml/datasets/Electricity Load Diagrams20112014
Dataset Splits No For the Electric dataset, the paper states '| (A, b)train| = 320, | (A, b)test| = 80'. Table A.1 lists 'Ntrain' and 'Ntest' for other datasets. However, no explicit details regarding a 'validation' split (e.g., percentages or sample counts) are provided for any dataset.
Hardware Specification Yes For the greedy method, we used several Nvidia Ge Force GTX 1080 Ti machines. For the maximum inner product method, the experiments are conducted on a laptop with a 1.90GHz CPU and 16GB RAM. The offline training is done separately using a single GPU.
Software Dependencies No The paper mentions using 'PyTorch (Paszke et al., 2019)' and refers to a 'package released in Agrawal et al. (2019)'. However, specific version numbers for these software components are not provided.
Experiment Setup Yes Experimental parameters (i.e., learning rate for gradient descent) can be found in Appendix G. For a given m, the dimensions of the four sketches were: S Rm n, R Rm d, S2 R5m n, R2 R5m d. Parameters of the algorithm: bs = 1, lr = 1.0, 10.0 for hyper and video respectively, num_it = 1000. ... In our experiments, we set bs = 20, iter = 1000 for all datasets. We set lr = 0.1 for the Electric dataset.