Stockfish, one of the best modern chess engines, is orders of magnitude stronger than DeepBlue.
Most chess engines have the same blueprint: abstracting a chess position, finding all candidates moves, iterating through the tree of candidates until a given depth and assessing the relevance of those moves to find the best one. Their quality, however, is subject to the following criterias :
- Speed: research of candidates moves and iteration of the tree,
- Accuracy : evaluation of the board.
This post will carefully examine the inner workings of the open-source chess engine Stockfish 12 and explain the key to its success. …
The quiet statisticians have changed our world; not by discovering new facts or technical developments, but by changing the ways that we reason, experiment and form our opinions. –Ian Hacking
Can you write a predictive function from observations which would minimize the error deviation? Or the average error?
Can you find the best theoretical predictor which would minimize any given loss function?
This post will answer those questions by providing a quick introduction to statistical decision theory. This is not another tutorial on the basics of linear algebra for machine learning. I assume a small mathematical background from the reader…
A practical application for a constraint solver. Studying other technologies can save you several days of data cleansing and model training.
Machine Learning and Deep Learning are ongoing buzzwords in the industry. Branding ahead of functionalities led to Deep Learning being overused in many artificial intelligence applications.
This post will provide a quick grasp at constraint satisfaction, a powerful yet underused approach which can tackle a large number of problems in AI and other areas of computer science, from logistics and scheduling to temporal reasoning and graph problems.
Solving a real-world problem
Let’s consider a factual and highly topical problem.
A pandemic is rising. Hospitals must organize quickly to treat ill people.
The world needs an algorithm which matches infected people and hospitals together given…
When viral marketing goes too far
A n Australian PhD candidate in artificial intelligence made a recent post on LinkedIn about his researches on SARS-CoV-2. The post gathered thousands of views, likes, and shares.
AI is wonderful
He built a Deep Learning model which is able to predict whether a patient is infected with the COVID-19 virus or not from chest radiographs with a 97.5% accuracy.
As it stands, the project features:
- A trained PyTorch model
- Containerized applicative code
- A GitHub repository translated in 8 languages
- A web application in development
- A mobile application in development
- Blueprints for a serverless architecture in AWS to host the model
- Lots of…
