aliquote.org

February 25, 2020

Dynamic Kernel Matching is analogous to a convolutional network, but for sequences.

DKM is analogous to a convolutional network, but for sequences. Consider the problem of classifying a sequence. Because some sequences are longer than others, the number of features is irregular. Given a specific sequence, the challenge is to determine the appropriate permutation of features with weights, allowing us to run the features through the statistical classifier to generate a prediction.