2: data2vec

<p>Todays paper: data2vec (https://arxiv.org/abs/2202.03555)<br/><br/><b>Summary of the paper</b><br/>A multimodal SSL algorithm that predicts latent representation of different types of input.</p><p><b>Highlights of discussion</b></p><ul><li>What are the motivations of SSL and multimodal</li><li>How does the student teacher learning work?</li><li>What are similarities and differences between ViT, BYOL, and Reinforcement Learning algorithms.</li></ul>