Turing complete neural computation based on synaptic plasticity

Area of Science:

• Neuroscience
• Computational Neuroscience
• Artificial Intelligence

Background:

• Neural computation traditionally encodes information in neuronal spiking, activation, or dynamics.
• Synapses and plasticity are primarily viewed as processing and learning mechanisms.

Purpose of the Study:

• To propose a novel Turing-complete paradigm for neural computation.
• To demonstrate that essential information can be encoded in discrete synaptic states.
• To explore the role of synaptic plasticity in information encoding and updating.

Main Methods:

• Developed a theoretical model of neural computation based on synaptic states.
• Utilized rational-weighted recurrent neural networks.
• Employed spike-timing-dependent plasticity (STDP) rules for information updating.
• Proved simulation of 2-counter machines and Turing machines.

Main Results:

• Demonstrated that discrete synaptic strengths can encode computational states and counter values.
• Showed that STDP rules can achieve transitions between synaptic weights.
• Established the theoretical possibility of Turing-complete synaptic-based neural computation.

Conclusions:

• Synapses can encode essential information, not just process it or facilitate learning.
• Synaptic plasticity mechanisms are capable of information updating.
• This synaptic-centric approach represents a paradigm shift in neural computation theory.