SMAUG: Sparse Masked Autoencoder for Efficient Video-Language Pre-training | IEEE Conference Publication | IEEE Xplore