Implicit Regularization in Tensor Factorization

Authors: Noam Razin, Asaf Maman, Nadav Cohen

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

Reproducibility Variable Result LLM Response
Research Type Experimental In this paper we provide the first theoretical analysis of implicit regularization in tensor factorization... Experiments validate our analysis, demonstrating implicit regularization towards low tensor rank in a wide array of configurations... we empirically explore its potential to serve as a measure of complexity for multivariable predictors.
Researcher Affiliation Academia 1Blavatnik School of Computer Science, Tel Aviv University, Israel.
Pseudocode No No pseudocode or algorithm blocks were found in the paper.
Open Source Code No The paper does not provide any concrete access information (e.g., a link or explicit statement of code release) for the source code.
Open Datasets Yes MNIST (Le Cun, 1998) and Fashion-MNIST (Xiao et al., 2017)
Dataset Splits Yes For each problem, we associate the label 1 with the active category and 0 with all the rest, and then attempt to fit training examples with predictors of low tensor rank, reporting the resulting mean squared error, i.e. the residual of the fit... We report this mean squared error, as well as that obtained by the predictor on the test set.
Hardware Specification No No specific hardware details (like CPU/GPU models, memory, or cloud instances) were mentioned in the paper for running experiments.
Software Dependencies No The paper cites PyTorch and scikit-learn, but does not specify the version numbers of these or any other software dependencies used for their experiments.
Experiment Setup Yes The first (left) three plots show (Frobenius) norms of the ten largest components under three standard deviations for initialization 0.05, 0.01, and 0.005... The tensor factorizations were initialized randomly with components drawn from a normal distribution with mean zero and standard deviation 0.01 (unless stated otherwise). Learning rates were kept constant, 0.01, (unless stated otherwise).