Spike-timing-dependent plasticity with weight dependence evoked from physical constraints

Area of Science:

• Neuroscience
• Computer Engineering
• Artificial Intelligence

Background:

• Spike-Timing-Dependent Plasticity (STDP) is a crucial learning rule in neural systems.
• Existing VLSI implementations of STDP face challenges in scalability and flexibility.
• Analog and mixed-signal Very-Large-Scale Integration (VLSI) offers potential for efficient neuromorphic hardware.

Purpose of the Study:

• To review existing analog and mixed-signal VLSI implementations of STDP.
• To present a novel, compact STDP circuit for large synaptic arrays.
• To demonstrate weight dependence and retention capabilities of the proposed circuit.

Main Methods:

• Developed a compact STDP circuit using silicon substrate limitations for weight dependence.
• Employed reverse-biased transistors to minimize leakage in weight-representing capacitances.
• Integrated the circuit for parallel use in large synaptic arrays.

Main Results:

• Demonstrated various methods for shaping the STDP learning rule.
• Showcased synaptic weight retention for several minutes.
• Observed shifts in synaptic weight distributions based on input correlational cues.

Conclusions:

• The presented compact STDP circuit is suitable for large-scale neuromorphic systems.
• The design offers flexibility in learning rule implementation and weight dynamics.
• This work contributes to the development of more efficient and biologically plausible artificial neural networks.