Convolution Nuclear Norm Minimization for Time Series Modeling
Introduce two recent papers about convolution nuclear norm minimization and how it works on time series data.
In the past few years, we witnessed the substantial progress in both matrix and tensor completion. The most important matrix completion model is built on nuclear norm minimization in which nuclear norm is the sum of singular values of matrix. This kind of matrix completion has been shown to be effective on real-world data of various types, ranging from images to time series. Today, I will give a brief introduction to the convolution nuclear norm minimization proposed in two recent papers [1–2]. Let’s get started!
Circular Convolution
As we know, convolution is a very important concept in functional analysis, which is also the cornerstone of many famous machine learning models (e.g., convolutional neural network (CNN)). Today, let’s first revisit what is the definition of circular convolution.
Note that our circular convolution might be different from some literature or posts because we just start the…
