Brain-Inspired In-Memory Computing
The BRICMEM project pushes the boundaries of in-memory computing for and beyond traditional machine learning applications by leveraging stochasticity, hyperdimensional representations, and the superposition principle. We are exploring this topic from all angles ranging from theory to experimental validation on existing and custom state-of-the-art hardware platforms.
Research Questions
BRICMEM draws inspiration from four key principles of biological information processing that could transform computational memory design.
- Can we leverage holographic vector representations and superposition to store information and compute more efficiently?
- Can we exploit intrinsic stochasticity to further enhance storage capacity?
- Can we communicate efficiently in time?
- Can we exploit the third spatial dimension to further enhance the storage capacity?
The group
We are a young and growing group and are continuously looking for new people and ideas.