ECE 50633 - Boltzmann Law: Physics to Machine Learning

Note:

This is a 5-week that corresponds to the end of ECE 50653.

Course Details

Lecture Hours: 3 Credits: 1

Counts as:

  • EE Elective
  • CMPE Special Content Elective

Normally Offered:

Each Fall, Spring

Campus/Online:

On-campus and online

Requisites:

MA 26200 or (MA 26500 and MA 26600)

Requisites by Topic:

Differential Equations and Linear Algebra

Catalog Description:

This course introduces the key concepts of equilibrium statistical mechanics leading to the celebrated Boltzmann law and how it leads to Boltzmann machines and related concepts in modern machine learning. No prior background in statistical mechanics is assumed.

Required Text(s):

  1. Lessons from Nanoelectronics, Part A: Basic Concepts (full text provided in Brightspace) , 2nd Edition , Datta, S. , World Scientific , 2017 , ISBN No. 13: 978-9813209749

Recommended Text(s):

None.

Learning Outcomes:

A student who successfully fulfills the course requirements will have demonstrated an ability to:
  1. Explain the law of equilibrium and entropy. [None]
  2. Explain how Boltzmann machines operate.. [1]
  3. Explain how quantum circuits operate.. [1]

Lecture Outline:

Weeks Topic
Week 1 Boltzmann Law: State Space, Boltzmann Law, Shannon Entropy, Free Energy, Self-Consistent Field
Week 2 Boltzmann Machines: Sampling, Orchestrating Interactions, Optimization, Inference, Learning
Week 3 Transition Matrix: Markov Chain Monte Carlo, Gibbs Sampling, Sequential versus Simultaneous, Bayesian Networks, Feynman Paths
Week 4 Quantum Boltzmann Law: Quantum Spins, One Q-Bit System, Spin-Spin Interactions, Two Q-Bit Systems, Quantum Annealing
Week 5 Quantum Transition Matrix: Adiabatic To Gated Computing, Hadamard Gates, Grover's Search, Shor's Algorithm, Feynman Paths

Assessment Method:

Students in this course will be evaluated by exams. (3/2022)