Learning Maximum Margin Temporal Warping for Action Recognition | IEEE Conference Publication | IEEE Xplore