Categories
Articles

Top Twitter Topics by Data Scientists #43

Trending this week: AlphaGo Zero explained in one picture; Rliable, an open-source Python library for reliable evaluation and comparison of DeepRL algorithms; Get the basics of the DL optimization theory.

Every week we analyze the most discussed topics on Twitter by Data Science & AI influencers.

The following topics, URLs, resources, and tweets have been automatically extracted using a topic modeling technique based on Sentence BERT, which we have enhanced to fit our use case.

Want to know more about the methodology used? Check out this article for more details, and find the codes in this Github repository!

Overview

In this new publication of our series of posts dedicated to the technology watch, we will talk about:

  • Some Crucial ML & DL Concepts
  • Amazing RL Resources
  • AI Achievements From 2021 & Trends For 2022

Discover what Data Science and AI influencers have been posted on Twitter this week in the following paragraphs.

Some Crucial ML & DL Concepts

This week, Data Science and AI influencers have shared some very interesting articles that introduce essential machine learning and deep learning concepts. They will help you to learn the basics of some aspects of ML and DL that are crucial to enable progress.

Image from Pixabay.

Mike Tamir has shared an introduction to the deep learning optimization theory. This blog post focuses on understanding the theory of optimization in the deep learning field, it aims to highlight interesting lessons learned from deep learning theory research that challenged conventional wisdom. It introduces the experimental and theoretical approaches to studying it.

On their side, ipfconline have posted an article about Receptive Field Calculations for Convolutional Neural Networks. This article explores the interesting and sometimes overlooked concepts of receptive fields that are related to convolutional neural networks — one of the most used deep learning concepts, used in a variety of industries for object detection, pose estimation, and image classification. Here, the intuition and math behind receptive fields are provided, as well as the implementation in Python of one version of a receptive field calculator.

Source: A guide to convolutional arithmetic Dumoulin et al. 2016

Finally, KDnuggets have shared a link to a book available on the internet for free that will give you an introduction to Machine Learning Interpretability. This guide provides you with cutting-edge interpretability techniques that will help you to improve the interpretability of your predictive models alongside their accuracy. Which will allow you to gain human trust, business adoption, and lower model validation efforts, and overcome regulatory oversight.

Amazing RL Resources

Some amazing posts on reinforcement learning (RL) have also been shared this week. Here is our selection:

First, KDnuggets have posted the 5 Things You Need to Know about Reinforcement Learning, one of the hottest research topics in artificial intelligence currently. This post is an RL 101 guide that will answer the following questions:

  • What is reinforcement learning? How does it relate to other ML techniques?
  • How to formulate a basic reinforcement Learning problem?
  • What are some most used Reinforcement Learning algorithms?
  • What are the practical applications of Reinforcement Learning?
  • How can I get started with Reinforcement Learning?

On his side, Mike Tamir has shared the following resources:

Reinforcement Learning Lecture Series 2021, taught by DeepMind researchers in collaboration with University College London (UCL) to offer students a comprehensive introduction to modern reinforcement learning. This series comprises 13 lectures, the series covers the fundamentals of reinforcement learning and planning in sequential decision problems, before progressing to more advanced topics and modern deep RL algorithms.

A post introducing the amazing tool Rliable, an open-source Python library for reliable evaluation, even with a handful of runs, on reinforcement learning and machine learnings benchmarks, with a visual explanation. Rliable was introduced by Deep RL at the Edge of the Statistical Precipice (Neurips Oral) to provide RL practitioners with a more rigorous evaluation and comparison of DeepRL algorithms results.

In this post, you will see the different tools used by rliable to better evaluate RL algorithms:

  • score normalization to aggregate scores across tasks
  • stratified bootstrap to provide proper confidence intervals
  • interquartile mean (IQM) to summarize benchmark performance
  • performance profile for an overview of the results and their variability
  • probability of improvement to compare two algorithms.
Performance Profiles — The proportion of runs that achieve target performance (Source: Image from the authors of the rliable library).

Finally, Kirk Borne has shared a more advanced post that explains AlphaGo Zero in one picture. AlphaGo Zero was announced by Google DeepMind in 2018. This extraordinary achievement has shown how it is possible to use deep reinforcement learning to train an agent to a superhuman level in the highly complex and challenging domain of Go, “tabula rasa”. The following picture shows how AlphaGo Zero is trained.

AI Achievements From 2021 & Trends For 2022

Finally, this week, some tweets on the AI achievements from 2021 and trends for 2022 have been posted.

Marcus Borba has shared an article on 7 achievements from 2021 driven entirely by AI. Including, fighting loneliness with AI and Google’s AI flood warning system in India

Kirk Borne has shared a tweet mentioning that he is honored to be included in the Main 2021 Developments and Key 2022 Trends in AI, Data Science, Machine Learning Technology.

Image from KDNuggets.

Finally, KDnuggets has shared the AI, Analytics, Machine Learning, Data Science, Deep Learning Research Main Developments in 2021 and Key Trends for 2022.

It included trending topics like — Sustainable AI, Explainable AI, Synthetic AI, No/Low Code AI, On-Device AI, Mission Critical AI, Regulation in AI.

Hope you enjoyed this new post of our technology watch series of articles. Stay tuned!