Sharp Recovery Thresholds of Tensor PCA Spectral Algorithms

Authors: Michael Feldman, David Donoho

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

Reproducibility Variable Result LLM Response
Research Type Experimental Even for modest-sized tensors, simulations demonstrate close agreement with theory. Figure 1: Simulations of tensor unfolding with k = 3, supersymmetric signal v1 = v2 = v3, and n1 {400, 800, 1200}. Solid lines display empirical cosine similarities (each point is the average of 50 realizations).
Researcher Affiliation Academia David L. Donoho Department of Statistics Stanford University donoho@stanford.edu Michael J. Feldman Department of Statistics Stanford University feldman6@stanford.edu
Pseudocode Yes Algorithm 1 Tensor unfolding; Algorithm 2 Partial tracing; Algorithm 3 Power iteration; Algorithm 4 Recursive unfolding
Open Source Code No The paper does not contain any statements about making its source code publicly available or providing links to a code repository.
Open Datasets No The paper uses synthetic data generated through simulations rather than a publicly available dataset. For example, Figure 1 describes "Simulations of tensor unfolding with k = 3, supersymmetric signal v1 = v2 = v3, and n1 {400, 800, 1200}."
Dataset Splits No The paper describes generating synthetic data for simulations and analyzing empirical results (e.g., "each point is the average of 50 realizations"), but it does not mention traditional training, validation, or test dataset splits.
Hardware Specification No The paper does not provide any specific details about the hardware (e.g., CPU, GPU models, or cloud computing resources) used to run its simulations.
Software Dependencies No The paper does not provide specific names or version numbers for any software libraries, frameworks, or tools used in its simulations or analysis.
Experiment Setup Yes The paper explicitly details the parameters for its simulations, such as in figure captions: "Simulations of tensor unfolding with k = 3, supersymmetric signal v1 = v2 = v3, and n1 {400, 800, 1200}." and "Simulations of tensor unfolding with k = 3, n1 = n2 = 50, and n3 = 10000 (left) or n3 = 20000 (right)."