Innovative Computing for AI

What it is about

The module conveys the limits of technological scaling and shows which novel concepts in hardware, memory and architecture are needed to overcome these limits. By the end you will know current Beyond-CMOS technologies, novel memory, in-memory/near-memory and neuromorphic approaches as well as brain-inspired algorithms (e.g., Hyperdimensional Computing) and their implications for performance, energy and reliability.

What you will be able to do

• Understanding the fundamental limits of technology scaling
• Knowledge of novel concepts beyond von Neumann architectures
• Familiarity with Emerging Logic- and Memory-Technologies
• Understanding reliability and endurance challenges in nano-technologies
• Knowledge of In-Memory, Near-Memory and Neuromorphic Computing
• Understanding brain-inspired algorithms such as Hyperdimensional Computing
• Combining algorithms and architectures for energy-efficient systems
• Insight into advanced thermal management and cooling

What the module consists of

• VorlesungConveying the fundamental concepts, discussion of scaling problems and presentation of current solutions from research and industry
• Übung/TutorialDemonstration of techniques, discussion of challenges, joint evaluation of pros and cons, work on small examples and paper discussions

Teaching method

• Lecture with presentations and board workExplanation of fundamental concepts, presentation of problems and state-of-the-art solutions
• Discussions and paper reviewsCritical engagement with research results and anticipation of possible solutions
• Tutorials with practical examplesPractice of techniques, discussion of implementation details and challenges