Hardware for Machine Learning

PhD Course January-March 2020, 6 credits

Machine learning (ML) has become an important tool to extract meaningful information from the vast data the pervasive sensors surrounding us generates (petabytes/s worldwide according to Cisco). However, with so much data to process comes very high computational demands on the hardware. This course aims at providing knowledge in different hardware solutions so that we can make better trade-offs in flexibility vs. efficiency when we are designing our data processing systems.

The PhD course is based on invited guest lecturers that will share their expertise in their respective areas. We plan to meet on Mondays for 8 weeks starting in late January 2020 following the program below. Prospective students should sign up with the examiner Mark Vesterbacka by 23 January 2020. To pass the course you need to attend the lectures and do the associated homework assignments. Deviations from this are discussed with the examiner.

Program

Lecture 1) Introduction to AI/ML

Guest lecturer: Fredrik Heintz

Time: Monday 27 January 2020, 10:15-12:00

Place: Systemet

Presentations: CourseInfo.pdf, Introduction_AI_ML.pdf

 

Lecture 2) Hardware Accelerators and Programming for ML

Guest lecturer: Christoph Kessler

Time: Monday 3 February 2020, 10:15-12:00

Place: Systemet

Presentation: Accelerators_ML.pdf

 

Lecture 3) GPU Architecture and Architecture Dependent Algorithms

Guest lecturer: Ingemar Ragnemalm

Time: Monday 10 February 2020, 10:15-12:00

Place: Systemet

Presentation: GPU_Architecture.pdf

 

Lecture 4) FPGA for ML

Guest lecturers: Fredrik Medley and Josefin Ringenson, Veonner

Time: Monday 17 February 2020, 10:15-12:00

Place: Systemet

The presentation is based on the work: Efficiency of CNN on Heterogeneous Processing Devices

 

Lecture 5) Robust Real-Time Face Detection

(from an academic paper to a custom built ASIC accelerator in 2x45 minutes)

Guest lecturer: Anders Lloyd, Axis Communications

Time: Monday 24 February 2020, 13:15-15:00

Place: Systemet

Preparation: Robust Real-Time Face Detection

Presentation: Paper_to_ASIC.pdf

 

Lecture 6) Spiking Networks

Guest lecturer: Robert Forchheimer

Time: Monday 2 March 2020, 10:15-12:00

Place: Systemet

Presentation: Spiking_Networks.pdf

 

Lecture 7) Near-Sensor Image Processing

Guest lecturer: Jörgen Ahlberg

Time: Monday 9 March 2020, 13:15-15:00

Place: Systemet

Presentation: Near_Sensor.pdf

 

Lecture 8) In-Memory Computation

Lecturer: Mark Vesterbacka

Time: Monday 16 March 2020, 10:15-12:00

Place: Systemet

Presentation: In-Memory_Compute.pdf

 

Informationsansvarig: Mark Vesterbacka
Senast uppdaterad: 2020-03-15