R tables, Robotics, and Reinforcement Learning!
Week 33-2023 [Issue 4.0]
Welcome to The Repo! 🚀
Each week, I'll curate three gems from the data science community:🗄️ Re: Remarkable Repository
💻 P: Prolific Programmer
🏢 O: Outstanding Organization
I hope you find them as valuable and insightful as I do.
You can find all recommendations in the GitHub repository at finnoh/repo!
rstudio/gt
🗄️ Repository | ggplot for R tables
For more extensive formatting of tables in R, that works with HTML, LaTex, and RTF formats, take a look at the gt R package. The philosophy behind this package reminds me a bit of the one behind ggplot, as each table is a collection of components. You can see the parts of a gt table in the figure below.
Eric Jang
💻 Programmer | Eric Jang tries to build machines that improve exponentially
Besides having been a Senior Research Scientist at Google, and serving as a Vice President of AI at 1X Technologies, Eric Jang is the co-inventor of the Gumbel Softmax. His main focus is on bringing machine learning principles from domains such as computer vision or natural language processing to robotics. He has also written a book, AI is Good for You, which is available on his website.
His post Can LLMs Critique and Iterate on Their Outputs?, shines a light on a difference in capability between GPT-4 and GPT-3.5: The ability to self-critique (to some extent). He showcases that the ability of a GPT-4 model to catch its mistake when prompted, does not imply that the model can also correct its mistake. The correction of the mistake might also be a wrong answer.
OpenAI—Spinning Up
🏢 Organization | A course that kickstarts your journey into Reinforcement Learning
OpenAI’s course offers explanations for the math behind Reinforcement Learning, as well as code implementations, a repository of key papers in the field, and tips on how to start a research project in Reinforcement Learning. I also thought that the repositories rlworkgroup/garage and openai/baselines are worth a look, as they provide tools for starting your Reinforcement Learning project.