Back to search

IKTPLUSS-IKT og digital innovasjon

Hybrid Deep Learning Cellular Automata Reservoir

Alternative title: Hybrid Dyp Læring Cellulære Automater Reservoar

Awarded: NOK 7.9 mill.

The DeepCA project aims at investigating a new artificial intelligence method that gets inspiration from nature, and in particular cellular systems. This allows the possibility to connect networks of biological neurons (grown in a laboratory) to an artificial intelligence system, and explore the potentials of a hybrid biological and artificial intelligence system. The results from the DeepCA project have the potential to contribute building more powerful and efficient computers (by looking at how networks of biological neurons grow and communicate) and to identify methods to connect computers to biological neurons (with potential medical applications for neural-related diseases).

The ambitious research goal of the DeepCA project is to create a theoretical and experimental foundation for a novel hybrid deep learning paradigm based on cellular automata and biological neural networks, in order to bridge the gap between neuroscience and deep learning towards self-learning devices that are significantly more efficient than the state-of-the-art. The desired results have the potential for breakthroughs in novel substrates for machine learning (easily transferrable to hardware implementations), as well as direct medical applications. Current deep learning implementations are not easily transferrable to hardware devices for widespread adoption, e.g., to sensor devices and Internet of Things, due to the required computing power and complexity of the underlying architectures. Therefore, a different computing paradigm is needed. Investigating biological neural cultures information processing and dynamics could lead to better, more powerful, energy efficient implementations of deep learning systems. The main hypothesis in DeepCA is that deep architectures of cellular automata reservoir can significantly facilitate the implementation of recurrent neural networks and reservoir dynamics, in order to efficiently interface biological neural networks and implement learning through local interactions. DeepCA combines Norway's leading experts in cellular automata and unconventional computing at OsloMet and neuroscience/neuro-computation at NTNU, with internationally leading researchers in morphogenetic engineering at Manchester Metropolitan University and hyperdimensional computing at Luleå University of Technology, forming together a very strong multidisciplinary team to complete DeepCA's ambitious research tasks. The project manager, Dr. Stefano Nichele, is a young researcher with a proven track record and ability to propose, manage and complete research projects. Nichele is 36 years old and currently Associate Professor at Oslo Metropolitan University.

Publications from Cristin

No publications found

Funding scheme:

IKTPLUSS-IKT og digital innovasjon