Independent component analysis for multivariate functional data

Joni Virta,Bing Li,Klaus Nordhausen,Hannu Oja

DOI: https://doi.org/10.1016/j.jmva.2019.104568

2017-12-21

Abstract:We extend two methods of independent component analysis, fourth order blind identification and joint approximate diagonalization of eigen-matrices, to vector-valued functional data. Multivariate functional data occur naturally and frequently in modern applications, and extending independent component analysis to this setting allows us to distill important information from this type of data, going a step further than the functional principal component analysis. To allow the inversion of the covariance operator we make the assumption that the dependency between the component functions lies in a finite-dimensional subspace. In this subspace we define fourth cross-cumulant operators and use them to construct the two novel, Fisher consistent methods for solving the independent component problem for vector-valued functions. Both simulations and an application on a hand gesture data set show the usefulness and advantages of the proposed methods over functional principal component analysis.

Statistics Theory

What problem does this paper attempt to address?

### What problem does this paper attempt to solve?

The main goal of this paper is to extend the Independent Component Analysis (ICA) method to multivariate functional data. Specifically:

1. **Background and Challenges**:
- Independent Component Analysis is a classical method used to extract independent source signals from observed mixed signals.
- Multivariate functional data is becoming increasingly common in modern applications, such as Electroencephalogram (EEG) data, socio-economic time series data, etc.
- Traditional ICA methods are primarily designed for vector data, and how to apply them to multivariate functional data is a new challenge.

2. **Research Contributions**:
- The paper proposes two methods: Fourth-Order Blind Identification (FOBI) and Joint Approximate Diagonalization of Eigenmatrices (JADE), and extends them to multivariate functional data.
- These methods define fourth-order cross-cumulant operators and use these operators to construct two new, Fisher-consistent methods to solve the ICA problem for multivariate functional data.
- The paper assumes that dependencies exist in a finite-dimensional subspace and defines the fourth-order cross-cumulant operators within this subspace.

3. **Practical Applications**:
- The effectiveness of the proposed methods and their advantages over functional Principal Component Analysis are demonstrated through simulations and applications to a hand gesture dataset.

In summary, this paper aims to develop ICA methods suitable for multivariate functional data to better extract important information from such complex data.