1. "How to teach a machine to program automatically?" ——NEAT Learning” (How do we teach a machine to program itself? — Neat learning) by Murat Vurucu
In this paper, Murat Vurucu explains a new method called "NEAT" in simple terms and examples.
NEAT generates new neural networks by integrating existing neural networks according to specific rules, which are inspired by genetics.
The only problem with this article: the word "neat" is not a pun, and does not include "simple".
2. "A practical guide to tree based learning algorithms" by Sadanand Singh
Are you looking for a practical and detailed tree-based learning algorithm tutorial? Don't look for it. Here is an in-depth tutorial, comparable to an enhanced version of Udacy, with theory, color pictures and code examples here.
The author of this tutorial, Sadanand Singh, also wrote a similar tutorial on Support Vector Machines (SVM), which can be read through the link (https://sadanand-singh.github.io/posts/svmpython/).
3. "Imagine this: Creating new visual concepts by recombining familiar ones" (Imagine this: Creating new visual concepts by recombining familiar ones)
symbol → image image → symbol
"white" suitcase "hat, orange floor, magenta walls"
The latest research published by Deepmind announces new progress in the development of algorithms that can generalize their own learning.
In the study, the researchers mimicked the way infants learn—seeing objects with the help of verbal cues and deriving meaning from them. They successfully demonstrated generalization to some colors and objects.
We also found it particularly interesting that researchers used unsupervised learning in their architectures. This paper definitely complements Future of Deep Learning (6th paper in the list).
4. Computer Reads Body Language by Byron Spice
The way the machine-learning system works is that it uses a video camera to learn how to recognize body language in real time, and can recognize the body language of a dozen people with just a laptop.
Once, Microsoft launched Kinect, a somatosensory gaming device for the Xbox 360 entertainment platform, but such glory days seem to have long since come to an end.
Now, researchers at Carnegie Mellon University are using a Kinect-like precise body mapping to train their AI system. Its input is live images recorded by a camera, and more accurate data is used as ground truth.
The AI system managed to generalize correctly about its training, and the results were pretty good. The researchers have made their code available as open source. Before viewing the code, be sure to watch the linked video (address: https://www.youtube.com/watch?v=LrCO8QcXfAY)
5. "How checkers was solved" by Alexis C. Madrigal
Before the battle between Deep Blue and Garry Kasparov, and the battle between AlphaGo and Lee Sedol, the earliest man-machine battle was the battle between the world's top checkers master Marion Tinsley and the Chinook supercomputer. Alexis C. Madrigal described this man-machine battle in the early 1990s with great interest in this article.
The Atlantic article not only explores the psychology of programmers and chess players, but also sheds light on current advances in machine learning. In addition, this article is also of great benefit to the improvement of readers'literary literacy.
6. The future of deep learning by François Chollet
In this article, François Chollet explores how to bring machine learning models closer to artificial general intelligence (AGI).
This article continues the ideas he presented earlier in "The Limits of Deep Learning" (both articles are contained in the author's book "Deep Learning with Python").
The author puts forward some pragmatic views on this topic, and the work of Agile data engineers is getting closer and closer to the realization of business and economic value. For them, this is definitely a good article!
7. "Technical debt in machine learning" by Maksym Zavershynskyi
Maksym Zavershynskyi provides a short but insightful overview of how technical debt can arise in machine learning projects.
Maksym's rhetoric is somewhat exaggerated, but it is still very admirable that he chose to explore such a topic that is rarely mentioned, and Maksym's advice is also of great practical value.
8. AI is changing how we do science. Get a glimpse
In this article, the famous journal "Science Magazine" gives 5 application cases of machine learning and AI in science, covering the fields of physics, psychology, biology, astronomy and chemistry.
We can learn from it that AI technology really shines in cutting-edge scientific experiments.
9. "I have data, I need to understand the data, where should I start?" "(I have data. I need insights. Where do I start? ) by Rama Ramakrishnan
If you've worked in data science, you've probably encountered this problem: most people don't know where to start.
Rama Ramakrishnan believes that we should explore a business as a black box. If you are a data scientist, you should ask yourself what kind of data you want to see before you study the data.
10. "A list of artificial intelligence tools you can use today — for businesses" by Liam Hänel
If someone tells you that artificial intelligence is still a laboratory thing, you can show him this list.
At present, artificial intelligence is widely used in business, and Liam Hänel has selected hundreds of companies applying artificial intelligence from different fields. The list is robust and features high-quality utilities.