Paperwithcode - State-Of-The-Art (sota) is a great website that tracks the latest paper and code in machine learning by categories and specific tasks/themes e.g., computer vision - object detection, natrual language processing - questions answering, playing games - startcraft etc.
For each categories and tasks/theme, this website kept updating newer papers with a link for github so you can see theories and practical codes to understand how to implement the theory.
This is one of the largest categories with 617 tasks. This is a huge list of images or video related tasks from classic computer vision problems like object detection, image classification to relatively newer problems like style transfer, zero-shot learning, talking head generation. What a massive list it is…
Natural Language Processing
This is the second largest categories with 193 tasks such machine translation, semantic analysis, text summarization, grammatical error correction.
It is too rough to summarize other categories as 'others' but audio, speech, game, robots, knowledge base, and adversarial are some of the examples.
As a weekly theme, it would be interesting to go deeper some of the categories and tasks that I might be interesting. For example, I am interested in style transfer, which can transfer the way of painting to other images. It means we can create a photo that I just take around me and convert it into a sort of paints as if Gogh or Picasso paints.