AI for Code and Science - A Neurosymbolic Programming Approach

Overview

Save Big on Coursera Plus. 7,000+ courses at $160 off. Limited Time Only!

Grab it

Explore the cutting-edge field of AI for code and science in this 43-minute tutorial session. Delve into the intersection of machine learning, programming languages, and software engineering, with a focus on blending symbolic and neural techniques. Learn about neurosymbolic programming and its applications in chemistry, physics, code repair, and neuroscience. Gain practical experience with artificial intelligence tools for developing scientific code. Discover the importance of interpretability in science and how scientific knowledge can be represented as code. Examine the neurosymbolic programming pipeline and its technical aspects. Participate in hands-on sessions covering molecular generation, symbolic regression for orbital mechanics, program repair, and coding for neuroscience. Engage in group discussions and explore how computing can help us understand the world through code.

Syllabus

Intro
Organizers
The state of artificial intelligence
Deep neural networks
transformer abstracted
What is a computer program?
What is program synthesis?
Types of program synthesis
Dimensions in program synthesis
This tutorial: Blending deep learning and program synthesis a new approach to artificial intelligence
Meet neurosymbolic programming
Neurosymbolic programming for science
Why code? Enables Interpretability in science
Scientific knowledge is code
The intersection of programming languages and Machine learning
A neurosymbolic programming pipeline
Survey on Neurosymbolic programming technical overview
Chemistry: Molecular Generation (day 1: morning)
Physics: symbolic regression for rediscovering orbital mechanics (day 1: afternoon)
Program repair (day 2 morning)
Code for Neuroscience (day 2 afternoon)
Group discussion
Expeditions in computing: understanding the world trough code