[Paper Review] Dopamine modulation via memristive schematic

Max Talanov, E. Yu. Zykov+3 more|arXiv (Cornell University)|Sep 19, 2017

Advanced Memory and Neural Computing3 references3 citations

TL;DR

This paper presents a memristive neuromorphic circuit that emulates dopamine-modulated synaptic plasticity using polyaniline-based memristors, implementing both excitatory (STDP) and inhibitory (iSTDP) learning with dopamine-like modulation via a tunable potentiometer. The key contribution is the physical realization and simulation of dopamine modulation in both excitatory and inhibitory plasticity, demonstrating scalable, biologically inspired learning with dynamic modulation of synaptic strength through neuromodulatory signals.

ABSTRACT

In this technical report we present novel results of the dopamine neuromodulation inspired modulation of a polyaniline (PANI) memristive device excitatory learning STDP. Results presented in this work are of two experiments setup computer simulation and physical prototype experiments. We present physical prototype of inhibitory learning or iSTDP as well as the results of iSTDP learning.

Motivation & Objective

• To develop a physical and simulated memristive circuit that emulates dopamine-modulated synaptic plasticity in artificial neurons.
• To implement both excitatory (STDP) and inhibitory (iSTDP) synaptic plasticity using operational amplifiers and memristive devices.
• To explore the effects of dopamine-like modulation on synaptic weight changes in both excitatory and inhibitory pathways.
• To validate the neuromodulatory function using a tunable potentiometer as a proxy for dopamine concentration.
• To investigate the feasibility of applying dopamine modulation to inhibitory GABAergic synapses, a previously underexplored area in neuromorphic systems.

Proposed method

• The system uses a three-part block diagram: excitatory (orange), inhibitory (blue), and modulatory (green) components, with op-amps implementing STDP and iSTDP learning rules.
• Hebbian STDP is implemented via op-amps U9–U11, while iSTDP uses U1, U7, U8, with time-locked pre- and post-synaptic spikes generating learning impulses.
• The modulatory function is implemented via a variable potentiometer (V(mod)) that controls the amplitude of LTP and LTD pulses, simulating dopamine's influence on synaptic strength.
• Memristive device conductivity is updated based on the time lag (Δt) between pre- and post-synaptic spikes, following the STDP rule Δw = 1/Δt.
• A multivibrator (U3) generates short output pulses when the integrated voltage exceeds a threshold, triggering synaptic weight updates.
• The physical prototype uses a wiring schematic (Fig. 2) and is validated through both simulation and hardware experiments with varying dopamine-like modulation levels.

Experimental results

Research questions

• RQ1 Can a physical memristive circuit accurately emulate dopamine-modulated STDP in excitatory synapses using a tunable potentiometer as a neuromodulator?
• RQ2 How does dopamine-like modulation affect the amplitude and timing of synaptic weight changes in both excitatory and inhibitory plasticity?
• RQ3 Is it feasible to extend dopamine modulation to inhibitory (GABAergic) synapses, as modeled by iSTDP, and what are the observable effects on learning dynamics?
• RQ4 Can the memristive device's conductivity be dynamically adjusted in response to neuromodulatory input, reflecting biological synaptic plasticity?
• RQ5 What is the relationship between the potentiometer setting (V(mod)) and the resulting amplitude of learning impulses in both STDP and iSTDP configurations?

Key findings

• The simulation results show that increasing the dopamine-like modulation level (via V(mod)) leads to a proportional increase in the amplitude of learning impulses, as seen in the green graph of Figure 4.
• In the physical implementation, the amplitude of STDP learning impulses increased from 0/50kΩ to 50/0kΩ potentiometer setting, confirming effective modulation of synaptic plasticity.
• The iSTDP learning function was successfully implemented and validated, with increasing modulation levels (0/1MΩ to 750/250kΩ) resulting in higher-amplitude learning impulses.
• The memristive device's overall conductivity changed in response to modulated learning impulses, as shown in the lilac curve of Figure 4, demonstrating effective weight update.
• The system successfully emulated both excitatory and inhibitory plasticity using operational amplifier-based circuits, with clear separation of LTP and LTD functions.
• The study provides the first physical realization of dopamine-modulated iSTDP, suggesting a pathway for future research into neuromodulation of inhibitory synapses.