Neuromorphic Machine Intelligence

About

The NMI Lab strives to create computing systems that work more like the brain, by combining the fundamental and the applied. Our long-term research goal is to emulate high-level, cognitive function in dedicated neuromorphic hardware systems.

The lab’s approach is to study neural circuits as computational substrates implementing a class of machine learning algorithms, and to design brain-inspired (neuromorphic) architectures that efficiently implement these algorithms.

Outcomes from this research range from brain-computer interfaces to goal-directed and adapting robotic systems.

Research Topics

  • Hardware-aware NeuroAI: Taking inspiration from the brain to build more efficient AI systems, with applications in vision, language and signal processing systems.
  • Physics- and Brain-inspired Learning Algorithms: develop learning and processing algorithms that can be implemented in existing and emerging neuromorphic hardware. Training remains the mostly costly operation in terms of energy and compute in the life time of a neural network. This severely limits the opportunity to specialize a network to a particular task, context or individual. We focus here on concepts of local learning: on a physical substrate, any computation ischaracterized by the set of variables available to the physical processing elements. Approaches explored in this research area include local self-supervised learning and contrastive learning, predictive codingand surrogate gradient methods.
  • Meta-Optimization for Hardware-Algorithm Co-design: Co-design designates the conjunctive development of algorithms and hardware. It is central tothe design of neuromorphic systems and the deployement of emerging materials and devices. A key vision of PGI-15 is to build a hardware-software-algorithm co-design ecosystem that replicates the multiple levels of learning and adaptation in natural organisms. We aim to introduce these multiple levels of learning to inject such inductive biases into the design, programming, and adaptation dynamics of neuromorphic hardware.

    • Contact

    Prof. Dr. Emre Neftci

    PGI-15

    Building TZA Aachen Aachen

    +49 241/92-780921

    E-Mail

    Members

    News
    No results found.
    Loading

    Last Modified: 21.05.2025
    #uk-e-rl-portal#uk-e-pof-fb-information-nacip-t3