Novel Approaches to Neural Computation

Dear Colleagues,

Artificial Neural Networks (ANNs) and their capability to address complex tasks with high efficiency when implemented on hardware have attracted remarkable interest for more than a decade now. Neuron models used in ANNs are highly simplified in function and operation to facilitate scaling to large networks. The most popular implementation that is widely adopted in modern AI hardware is the Rectified Linear Unit (ReLU). Such a simplified model is sufficient to train large networks with the backpropagation algorithm and to yield exceptional accuracy, though it is not a biomimetic neuron model since it does not generate 'spikes' and neglects the highly nonlinear nature of neurons and their temporal dynamics. On the other hand, Neuromorphic Spiking Neural Networks (SNNs), which are bio-inspired, use a Leaky Integrate-and-Fire (LIF) neuron which can generate voltage spikes or ‘events’ similar to biological neurons. The LIF neuron realizes the step/sigmoid activation, using rate coding of spikes, which is also a simple form of nonlinearity generating ‘all-or-nothing’ events as voltage spikes. SNNs mimic biology better than ANNs but still fail to capture the temporal dynamics of biological neurons to exhibit more complex behavior such as burst generation. Though ReLU and LIF neurons have been extremely successful when used in a predominantly feed-forward network, they are not suitable for demonstrating neuroscientific principles at the nodal and network levels, such as multiple modes of operation (spiking, bursting, etc.), coupled oscillations with excitatory and inhibitory connections, and rhythm generation with neurons in-phase or out-of-phase.

Biological neurons exhibit rich nonlinear dynamics both at the single neuron level and at the network level, which is enabled through neuromodulation using multiple feedback paths (local and global) operating at multiple timescales. Complex neuron models inspired by neurophysiology have been proposed before, like the Hodgkin–Huxley model and the Izhikevich model. These models capture the biophysics of the neuron accurately but are based on non-intuitive computationally expensive differential equations that are hard to implement on hardware with minimal circuit elements.

This Special Issue will focus on “Novel Approaches to Neural Computation”, and we invite innovative and breakthrough ideas on implementing neural computation that goes beyond the limitations of the current SNN architectures to achieve features/functions not possible before to perform new tasks and/or improve the performance in existing applications and thereby pushing the boundaries of what is possible.

Topics of interest include but are not limited to the following:

• Novel hardware-friendly neuron models, not limited to biophysical or biomimetic topologies;
• Innovative approaches to computing with spikes or events, especially using spike times or combining rate/time coding of spikes;
• Network architectures that use the potential of bursting neurons to demonstrate oscillations and rhythms;
• Applications and benchmarking using novel neural computation principles.

Dr. Rajkumar Kubendran
Guest Editor

Manuscript Submission Information

Manuscripts should be submitted online at www.mdpi.com by registering and logging in to this website. Once you are registered, click here to go to the submission form. Manuscripts can be submitted until the deadline. All submissions that pass pre-check are peer-reviewed. Accepted papers will be published continuously in the journal (as soon as accepted) and will be listed together on the special issue website. Research articles, review articles as well as short communications are invited. For planned papers, a title and short abstract (about 100 words) can be sent to the Editorial Office for announcement on this website.

Submitted manuscripts should not have been published previously, nor be under consideration for publication elsewhere (except conference proceedings papers). All manuscripts are thoroughly refereed through a single-blind peer-review process. A guide for authors and other relevant information for submission of manuscripts is available on the Instructions for Authors page. Electronics is an international peer-reviewed open access semimonthly journal published by MDPI.

Please visit the Instructions for Authors page before submitting a manuscript. The Article Processing Charge (APC) for publication in this open access journal is 2400 CHF (Swiss Francs). Submitted papers should be well formatted and use good English. Authors may use MDPI's English editing service prior to publication or during author revisions.

• neuromorphic
• spike-based computing
• biomimetic
• bio-inspired