Master Thesis Exploring Formal Guarantees for Neural Network Optimizations
Core
Developing a mathematical framework to formally verify numerical tolerance limits for neural network operations and evaluating trade-offs between hardware acceleration and numerical stability for efficient deployment on edge/embedded hardware.
Role type
Master Thesis Researcher (Neural Network Verification & Optimization)
Builds
Formal verification metrics and optimized neural network deployment strategies for HW/SW co-design
Domain
Artificial Intelligence / Embedded Systems / Hardware-Software Co-design
Deliverable
research
Required skills
Python, C++, Neural Network Compiler (TVM), Deep Learning Concepts, Linux, LaTeX
Preferred skills
Basic hardware knowledge, Inventive spirit
Technologies
TVM, Python, C++, Linux, LaTeX
Responsibilities
Analyze state-of-the-art optimization strategies for neural network deployment, develop a mathematical framework for formal verification, evaluate interplay between hardware acceleration and numerical stability, document research findings and present to the team
Seniority
Master Thesis Candidate
