• Print
• View
• History
• Login

Artificial Intelligence

Artificial Intelligence

(:toc:)

Artificial Intelligence - Learning Materials

• Favourite Quick Start Course material : Deep Learning Indaba
  • Nice presentation to understand the mathematics behind CNN training Mathematics for Machine Learning slides
• Text book - Deep Learning or Deep Learning ... according to Elon Musk "Written by three experts in the field, Deep Learning is the only comprehensive book on the subject".
• Quickstart tutorial
  • Look at the Keras and Tensorflow example
    • Test python program that creates a CNN to classify digits in the MNIST data set.
    • test.py
  • Nice cover article for this Image Classification in 10 Minutes with MNIST Dataset
  • Another nice cover article for this Keras Tutorial: The Ultimate Beginner’s Guide to Deep Learning in Python
• Mitchell's AI Workshop
  • Learn AI
• Kyle Cranmer's Course
  • Statistics and Data Science Jupyter Book
• Harvard CS50’s Artificial Intelligence with Python – Full University Course Foundational Concepts and Algorithms

Computing Resources

• Free Google GPU Works like Jupiter Notebook ... just create a new one (use menu bottom right)
  • Copy your code into the editor
  • Use the "edit" top level menu to set the "NoteBook Setting" to use the GPU under the hardware accelerator
  • Press the "play" arrow icon
• Amazon Cloud
• Local GPU (access details private)

AI in High Energy Physics

• There is extensive use of machine learning at many levels in the analysis of ATLAS data
• Public page of AI in ATLAS
• CERN Courier - The rise of deep learning
• Paper : Deep Learning and Its Application to LHC Physics
• Slides : Machine Learning @ ATLAS

AI for MinPET

• MinPET is the All-Machine discovery of diamond in kimberlite rocks using PET.
  • MinPET website
  • Recent Paper in the 2017 SAIP proceedings
  • MinPET in the CERN Courier
  • We have used AI to achieve diamond classification from PET images at speeds that enable real time run-of-mine rock sorting at many hundreds of tons per hour.
  • Commercialisation of the Technology prevents further exposition - If you are interested, contact the Research Team.