Some links and papers that I have found interesting this week. If you have any comments, please let me know.
1: Perspectives on the State and Future of Deep Learning - 2023, a very good paper that summarizes the current state of deep learning and the challenges that it faces in the future. It is a long read, but it is worth it. They interview Kyunghyun Cho, Zachary C Lipton, Melanie Mitchell, Preetum Nakkiran, Max Welling, Andrew Gordon Wilson on five pressing questions about the future of deep learning.
2: Deep Learning Tuning Playbook is a very good resource for maximizing the performance of deep learning models. It is a collection of best practices and tips for tuning deep learning models.
3: Choose Your Weapon: Survival Strategies for Depressed AI Academics by Julian Togelius and Georgios N. Yannakakis is a kind of satirical paper about the current state of AI research and what small labs can do to stay competitive while remanining academic. It is a very good read.
4: A PhD in Numbers by David Stutz is a very good post about his journey through his PhD. He talks about his experience, the challenges he faced and the lessons he learned and what he accomlished each year.
5: Role-Playing Paper-Reading Seminars by Alec Jacobson and Colin Raffel. I just found this method of running a paper reading seminar very interesting. I have only partipated in regular seminar during my PhD, but I think this method could be very useful to get more out of the papers.
6: The Essense of Global Convolution Models by Ben Athiwaratkun is a very good technical post about global convolution models. He explains the intuition behind them and how they work and help me understam the new Mamba model. He also provides a very good list of references.
7: Some papers I read this week and found interesting: Exploiting Inductive Biases in Video Modeling through Neural CDEs, Quantifying & Modeling Multimodal Interactions:
An Information Decomposition Framework, To Compress or Not to Compress - Self-Supervised Learning and Information Theory: A Review
comments powered by