A-ViT: Adaptive Tokens for Efficient Vision Transformer | IEEE Conference Publication | IEEE Xplore