A new artificial dendritic neural circuit inspired by the brain

With rapid advances in artificial intelligence (AI) tools, engineers around the world are working on new architectures and hardware components that replicate the organization and functions of the human brain.

Most brain-inspired technologies created to date are inspired by the firing of brain cells (i.e. neurons), rather than reflecting the overall structure of neuronal elements and how they contribute to information processing.

Researchers at Tsinghua University recently presented a new neuromorphic computational architecture designed to reproduce the organization of synapses (i.e., the connections between neurons) and the tree-like structure of dendrites (i.e., the projections extending from the body of neurons).

This new artificial brain-like system, presented in an article published in Natural electronicswas achieved using a computer model of multi-gate silicon nanowire transistors with ion-doped sol-gel films.

“When I was a master’s student in AI and brain bioengineering at Politecnico di Milano in Italy, I had the idea of ​​mimicking the sparseness and morphology of brain connectivity, such as that of neuron dendrites, to design efficient AI,” Carlo Vittorio Cannistraci, one of the corresponding authors, told Tech Xplore.

“I was also fascinated by the elegance of brain mechanisms such as ‘silent synapses’ that activate when surrounded by increased electrical activation.”

Drawing on his previous studies and interests, Cannistraci has recently set out to realize complex brain mechanisms through computational calculations. In this recent study, he teamed up with other researchers at Tsinghua University to reproduce the morphology of dendrites and the foundations of synapses using a neuromorphic computer model.

“One day, Carlo asked me to study ‘dendritic computing’ because our previous collaborative study on ‘neurotransistors’ had the potential to mimic dendritic properties,” Eunhye Baek, one of the corresponding authors, told Tech Xplore.

“Professor Luping Shi and I were looking for ways to develop a neuromorphic visual sensor system and we recognized the potential of this approach.

“My interest has always been in building dynamic information processing systems that are closer to those of the brain and neurons. Dendritic computation has really excited me because it encompasses a wide range of dynamic and complex properties that have not yet been studied in depth in neuromorphic engineering.”

Most research to date on neuromorphic computing has focused on reproducing synaptic processes associated with learning and artificially replicating neuronal spike generation. These studies have often modeled dendrites as simple transmission lines, ignoring the functions associated with their unique morphology.

“Dendrites use their tree-like morphology to map spatially distributed signals, exhibiting branch-specific plasticity and integrating diverse synapses,” Baek explained.

“Each dendritic branch is particularly sensitive to signals with a specific directionality, making them specialized in processing spatiotemporal signals. Our research focuses on these complex dendritic functions.”

Cannistraci, Shi, Baek and their collaborators have designed and developed a new device that mimics the morphology and function of biological dendrites. The device, called a dendristor, mimics the computations performed by dendrites by exploiting the physics of multi-gate transistors coated with an ion-doped sol-gel film.

“This film mimics dendritic branches by allowing doped ions to move in a manner similar to ions in neuronal dendrites, modulating the transistor current to reflect changes in dendritic membrane potential,” Baek said. “Our study demonstrates that the dendritor exhibits nonlinear dendritic integration and direction selectivity.”

In addition to the dendristor device, this research group’s recent paper presents an artificial silent synapse. In this system, the tension of the dendristor branches in the sol-gel film ensures that synaptic inputs only activate when the film reaches a specific threshold, thereby improving the system’s ability to discern the direction of moving visual stimuli.

“We also created a neuromorphic dendritic neural circuit that calculates the direction of moving signals, inspired by neural circuits in the retina and visual cortex,” Baek said. “This circuit shows the ability to detect signals moving in 2D and depth, integrating them to reconstruct the direction of motion of objects in 3D space.”

By faithfully reproducing the sparse connectivity of dendritic neurons, the novel neuromorphic computing approach introduced by Cannistraci, Baek, and their colleagues has achieved remarkable energy efficiencies. In fact, this system demonstrates the potential to detect motion using fewer neurons than existing artificial neural networks (ANNs).

The main advantage of this new architecture is that it goes beyond simply replicating the functional aspects of biological neurons. Unlike other existing neuromorphic computing platforms, it also reproduces the structure and sparse connectivity of neurons, including the morphology of dendrites and the foundations of silent synapses.

“Although there are different approaches in neuromorphic research to achieve intelligence, our study uniquely shows the importance of neuron morphology and their synaptic connections in dynamic signal processing,” Baek said.

“We achieved this by mimicking how biological neurons form functional neural circuits with spatially sparse mapping of synaptic inputs, highlighting how this morphology is crucial for efficient processing of neuromorphic information.”

This research team was notably the first to demonstrate that the spatial position of inhibitory and silent synapses can also control signal processing by neurons in neuromorphic systems. This discovery could guide the design of other computer models and architectures reproducing silent synapses.

“Sparsity and morphology have been poorly understood and misused to build next-generation AI,” Cannistraci said. “Our study is the first to show how to exploit these two characteristics of real brain networks to design next-generation neuromorphic neural networks for effective AI.”

The recent work by Cannistraci, Baek, and their colleagues could soon open up exciting new avenues for engineering neuromorphic systems based on semiconductor devices. Specifically, their proposed brain-inspired design could help develop new AI devices and tools that consume less energy, paving the way for more sustainable computing.

In their next studies, the researchers plan to further extend their artificial neural circuits, using advanced inhibitory connections that could further improve the classification of dynamic visual signals. To do this, they will attempt to closely mimic the neural connections observed in the brain at its earliest stages of development.

“We plan to develop new neuromorphic dendritic network architectures that perform deep learning and can solve other AI tasks beyond visual perception, such as time series analysis and auditory tasks,” Cannistraci added.

“Furthermore, we want to develop multimodal circuits capable of processing and correlating sensory inputs of different types, such as visual and acoustic. Finally, we want to extend this sparse and morphological computation paradigm to classical types of artificial neural networks implemented on digital hardware.”

© 2024 Science X Network