Full CMOS Analog Circuit Implementation of Multi-Functional Pavlov Associative Memory using STDP Learning

This paper demonstrates the complete Pavlov associative memory at the CMOS transistor level. Unlike many works focusing on digital implementation with area-intensive and power-hungry circuits, the proposed design for Pavlov associative memory is realised in the analog domain. Apart from learning and forgetting shown in various literature, the proposed circuit for Pavlov associative memory also incorporates the other biologically inspired functions of associative memory, such as learning with stimuli interval, variable learning rates, and generalisation and differentiation. Besides, most papers employ Pavlov training by forming a strong synaptic connection between the sensory auditory and output salivary neurons. In contrast, the actual Pavlov conditioning strengthens the synaptic connection between the auditory and gustatory neurons, as implemented in this paper. Unlike many works that use software-based SPICE models of a memristor as a synapse, the complete system uses on-chip trainable CMOS STDP memristive synapses and LIF neurons. The post-layout simulations of the proposed circuit implemented in TSMC 180 nm CMOS technology verify the functionality of the complete system.