Tensor Canonical Correlation Analysis

The code that accompanies the paper "Tensor Canonical Correlation Analysis withConvergence and Statistical Guarantees" by Chen, Kolar and Tsay.

Link: https://arxiv.org/abs/1906.05358

Basic Usage

We implement several algorithms in util.py

• twoDcca(X, Y) – 2DCCA for first left and right canonical component.
• twoDcca_mat(X, Y, p1, p2) - (p1,p2)-2DCCA for computing p1 left and p2 right canonical components.
• twoDcca_deflation(X, Y) - The deflation procedure for 2DCCA

Please see example_util.ipynb for more details.

Comparison

We compute TCCA for three datasets in Comparision.ipynb. The result is shown in Table 1 of our paper.

Reference: AppGrad: https://github.com/MaZhuang/CCA

rifle: https://cran.r-project.org/web/packages/rifle/index.html

Air Pollution Data in Taiwan

Original data can be download from https://taqm.epa.gov.tw/taqm/en/YearlyDataDownload.aspx. We produce the results in section 6.1.2 in twn_air_pollution.ipynb.