Deep InfoMax Tensorflow-Keras Implementation

Self-supervised representation learning

Learning a simpler representation of data has been an important application of deep learning. These simpler representations can be employed for a wide range of downstream tasks, or be used by other systems or neural networks. One of the most common examples is “word2vec” which represents every word via a 300 dimension vector. These vector representations have been widely employed in NLP applications and tasks, such as sentiment analysis, summarization, translation, etc.