Annotated Machine Learning Research Papers

Overwhelmed by the thought of reading research papers? Perhaps you feel there are too many to keep up with. Or maybe you dread trying to make sense of those you choose to move ahead with reading.

In either of these cases, you may find some annotation helpful. Sure, you could do it yourself and have this markup available for subsequent reading, but machine learning engineer Aakash Kumar Nain (@A_K_Nain) has made available a collection of machine learning research papers that they have annotated during reading and shared with the community at large.

From annotated version of "What is being transferred in transfer learning?"

Aakash explains the importance of reading research papers in general, and the importance of doing so for his own work.

I spend a lot of time reading papers. It is a crucial part of my ML work. If you want to do research or you want to be a better ML engineer, then you should read papers. This habit of reading papers will help you to remain updated with the field.

The focus here is on quality over quantity. The current modest number of papers in the collection are thoroughly annotated, making an investment of time in their reading well worth it.

As you can likely see, the results of these annotated papers are both informative and visually appealing.

From contrastive learning to meta-learning to CycleGAN to Transformers for image recognition, there is a variety in the selections. Here are links to the papers included at the time of publication:

• Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks
• Adaptive Risk Minimization: A Meta-Learning Approach for Tackling Group Shift
• Axial DeepLab: Stand-Alone Axial-Attention for Panoptic Segmentation
• What Should Not Be Contrastive in Contrastive Learning
• Unsupervised Learning of Visual Features by Contrasting Cluster Assignments
• An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale
• Flow-edge Guided Video Completion
• Supervised Contrastive Learning

From annotated version of "Supervised Contrastive Learning"

Whether you are new to the idea of reading machine learning research papers or someone who regularly indulges, this small collection of annotated papers may provide some useful insights when you next have free time.

 
Related:

• Papers with Code: A Fantastic GitHub Resource for Machine Learning
• AI Papers to Read in 2020
• Getting Started in AI Research