To analyze the abundance of multidimensional data, tensor-based frameworks have been developed. Traditional matrix-based frameworks extract the most relevant features of vectorized data using the matrix-SVD. However, we may lose crucial high-dimensional relationships in this process. To facilitate efficient multidimensional feature extraction, we propose a projection-based classification algorithm using the t-SVDM, a tensor-based extension of the matrix-SVD. We apply our algorithm to the StarPlus functional Magnetic Resonance Imaging (fMRI) dataset. Through our numerical experiments, we conclude that there exists a more accurate tensor-based approach to fMRI classification than the best possible equivalent matrix-based approach. Our research showcases the potential of tensor-based classification frameworks, and justifies further research into the usage of tensors for numerous other classification tasks.

Loading...

Taking too long?

Reload document

| Open in new tab