Author: admin

Our final NEUChip project meeting brought together partners from across Europe and beyond, including teams from Barcelona, Israel, Paris, Aston, Loughborough, and 3Brain, to share progress and discuss key results. Hosted in the historic Erasmus Darwin House on Beacon Street, just a stone’s throw from Lichfield Cathedral, the meeting provided an inspiring setting to exchange ideas and refine our collaborative research efforts. The discussions highlighted exciting advances in neuromorphic computing and strengthened our collective vision.

Beyond the formal sessions, the group enjoyed engaging and productive discussions, fostering deeper collaboration and shared enthusiasm for the research. In the evening, we gathered for a delightful meal, continuing conversations in a more relaxed setting. To round off the experience, we embraced the local atmosphere by joining in with traditional shanty songs at a nearby pub—an unexpected but thoroughly enjoyable highlight of the visit!

Talk given by Prof. Shahar Kvatinsky

Today’s AI applications demand tremendous computing power. However, existing hardware for AI is hitting a bottleneck in terms of speed and power. This is particularly because of the data movement required due to the separation of computing and memory in the von Neumann architecture and the large energy consumption, access time and cost of existing memory technologies. Existing AI uses devices like GPUs and dedicated hardware like TPUs and edge inference devices like ASICs. However, the brain is able to perform its processing at very low power and is particularly good at perception tasks.

In this talk, Prof. Shahar Kvatinsky explained potential ways to improve hardware for AI by brain inspired neuromorphic computing using emerging memory devices called memristors. Memristors can emulate synaptic functions and can be used to accelerate neural networks. One way is to perform Vector Matrix Multiplication (VMM) and Multiply and Accumulate (MAC) operations using crossbars of memristors. He explained training of memristive neuromorphic systems using backpropagation and stochastic gradient descent, methods for low power neuromorphic computing like low precision AI inference and trainable data converters and discussed some of the security issues and mitigation strategies. He also talked about various emerging memory devices that can be used to build these systems like Magnetic Tunnel Junction (MTJ) based devices and floating gate flash memory called Yflash, built by Tower Semiconductor. He described quantized deep neural networks with MTJ based devices and simpler neural network models like Deep Belief Networks (DBNs) built with Yflash devices.

Prof. Shahar Kvatinsky’s talk highlighted some of the devices and techniques to accelerate and improve the performance of hardware for AI applications.

Written by Rishona Daniels

field of biological machine learning, promising attempts have been made to use cortical neurons for machine learning tasks. However, these efforts often lack a strong theoretical foundation from a mathematical perspective.

Prof Saad’s talk introduced three mathematical tools to address this gap:

1. Neuronal Network Inference: Saad presented an algorithm that uses machine learning and statistical physics to infer the structure and connectivity of neuronal networks from spontaneous activities. This is crucial for understanding neuronal architecture and plasticity.
2. Visual Informatics Approach: He showcased methods to study differences in neuronal activities under various conditions using advanced dimensionality reduction techniques, retaining the global structure of high-dimensional data better than traditional methods like PCA and t-SNE.
3. Spatial Entropy Measurement: Saad emphasized using a Bayesian approach to measure the spatial entropy of neuronal activities accurately, revealing significant differences between spontaneous and stimulated activities. This provides a reliable metric for studying neuronal behaviour.

Overall, Prof Saad highlighted the potential of mathematical tools to enhance our understanding of neuronal networks and advance biological machine learning.

Presenting the consorta at the 2023 International Symposium on Nonlinear Theory and Its Applications (NOLTA) a number of the group were able to present over the week event, bring in lots of discussion and interest from others, as well as push forward ideas amongst NeU-ChiP partners in various places, even on Mount Etna. Papers and abstracts from the event can be found here.

A special symposia for NEU-ChiP ran over two days as a hybrid meeting, enabling discussion with collaborators around the world.

Around the physical sessions the NEU-ChiP team found time to discuss ideas.

Around the conference we had the chance to see a little of the city of Catania, with some great food, and an excellent conference dinner in the beautiful museum.

The NEU-ChiP team also found some time to enjoy the surrounding sites with some adventures up to Mount Etna.