ECE 386: AI Hardware Applications

This is the course website for ECE 386 - AI Hardware Applications, Spring ‘25.

Many of the pages are Jupyter Notebooks and can be opened in Google Colab by clicking the launch button 🚀 at the top right of the page!

Assignment Due Dates

All assignments are submitted via Gradescope.

Prog

Lesson Due	Type	Tasks
06	ICE	K-means classification of handwritten digits
07	ICE	DNN classification of handwritten digits
08	Prelab	Raspberry Pi setup, FastAPI demo from “Cloud Hosting”, System design
09	Lab	Handwritten Digits DNN & FastAPI
10	C&C	Block 1: Machine Prediction
11	Prelab	ECE 281 instruction set architecture review, Your questions about DSP reading, Arduino IDE setup
14	Prelab	Key word spotting (KWS)
16	Lab	ARM Cortex Benchmark Report
17	Lab	KWS
18	C&C	Block 2: Edge Inference
18	Prelab	Cats vs. Dogs

Schedule

Machine Prediction

• 1. Hello, Colab!
• 2. Networks and Tooling
• 3. Prediction Machines
• 4. Datasets
• 5. ICE 1: K-Means Clustering
• 6. ICE 2: Handwritten Digits - DNN
• 7. Cloud Hosting
• 8. Lab 1: Handwritten Digits - FastAPI
• 9. C&C: Prediction and Dimensionality

Edge Inference

• 10. Edge Inference Intro
• 11. Lab 2: Cortex DSP Benchmark
• 12. ARM Cortex Architecture
• 13. Quantization
• 14. Lab 3: Edge Impulse KWS
• 15. Lab Workday
• 16. Command and Risk
• 17. C&C: Edge Inference

Development Boards

• 18. Lab 4: Cats vs. Dogs
• 19. Lab Workday
• 20. 🐱 🐶 Live Inference
• 21. Containerization
• 22. ICE 3: Whisper Transcription
• 23. GPU Architecture
• 24. ICE 4: GPU Acceleration
• 25. Student Choice

Large Language Models

• 26. Text Vectorization
• 27. Large Language Models
• 28. Lab 5: Prompt Engineering
• 29. Lab Workday
• 30. C&C GPUs and LLMs

Final Block

• 31. Final Project
• 32. Final Project Workday 1
• 33. Final Project Workday 2
• 34. Final Project Workday 3
• 35. Final Project Workday 4
• 36. Wrap and Critique

Grading

This course has the following types of assignments:

• Prelabs are due before the corresponding lab starts

• In-class exercises (ICE) are designed to not take much more than one class period

• Labs require programming work outside of class as well as a report

• Daily briefs are based on the lesson reading or previous lesson content

• Final project is cumulative and worth 25% of the grade

        pie title Final Grade Weighting
    "Prelabs/ICE" : 25
    "Labs" : 30
    "Daily Briefs" : 15
    "Quizzes" : 30
    

        pie title Final Grade Weighting
    "Prelabs/ICE" : 20
    "Labs" : 25
    "Daily Briefs" : 5
    "Quizzes" : 20
    "Final Project": 25
    "Instructor Prerogative​"  : 5
    

Blackboard tabulates and holds official grades.

Course Info

ECE 386: AI Hardware Applications 3 (1) Students will use Artificial Intelligence pipelines and frameworks to train and deploy machine learning models onto various computational platforms. Deployments will consider engineering tradeoffs between model size, hardware constraints, and system design grounded in an understanding of computer architecture. Students will use Python to implement training on cloud-based hardware accelerators. Inference will be conducted on microprocessor and GPU boards. Emphasis will also be placed on the balance of AI provided prediction with human provided judgment. Lab. Final project or final exam. Prerq: Comp Sci 110; ECE 281 or Comp Sci 351 or department approval. Sem hrs: 3 fall.

Course Goals

Upon completion of this course, cadets will be able to 1) understand AI as a form of prediction that informs decision-making, 2) design and construct AI training pipelines, 3) deploy AI models on appropriate hardware for inference.​

Course Objectives

Each objective supports an ABET Student Outcome for accreditation.

1. Apply machine prediction to aid in making risk-based decisions in an uncertain or adversarial environment.​

   • #4 – Professional and ethical responsibilities; judgements considering societal factors​

2. Design and construct a machine learning training pipeline using contemporary frameworks and tools.​

   • #1 – Solve complex engr problems by applying engr, science, and math​

3. Evaluate the engineering tradeoffs between devices when deploying models for inference.​

   • #6 - Develop and conduct experiments; analyze data; draw conclusions​

4. Demonstrate the impact of computer architecture on AI computation.​

   • #2 – Engr Design to meet needs while considering societal factors​

Prerequisites

• CompSci 110 (for Python programming)

• ECE 281 or CompSci 351 or department approval (for computer architecture, ALU, datapaths)

Course Text Book

Books for this course are available for free from O’Reilly via DoD MWR Libraries.

Disclaimer

The contents of this website are for educational use only and do not necessarily reflect the official policy or position of the United States Air Force Academy, the Air Force, or the U.S. Government.