Tutorials

Neuromorphic Computing: Biological Foundations - This set of slides explains the ionic channels and gating mechanisms of biological neurons, ionic mechanisms behind the membrane resting potential, and Nernst and Goldman-Hodgkins-Katz potentials. These concepts are then used to develop a circuit model of the membrane, which is followed by discussions on electrical and chemical synaptic transmission mechanisms, generation and propagation of action potentials, and refractory periods. Ionic mechanisms underlying the generation of action potentials and establishment of refractory periods are also discussed.

The latter half of this set of notes (from slide 87 onward) discusses synaptic plasticity, a mechanism by which the neural activity generated by a stimulus alters the efficacy of neuronal communication via modification of the strength or synaptic transmission at preexisting synapses. Discussed in detail are the Hebbian rule of learning, long term potentiation/depression (LTP/LTD), and Spike Timing Dependent Plasticity (STDP) and its variants.

Neuromorphic Computing: Neuron Models - This set of slides discusses the most commonly used dynamical models of biological neurons. Included are discussions of the Hodgkin-Huxley model, Izhikevich model, leaky integrate and fire (LIF) model, generalized LIF model, nonlinear models such as Exponential Integrate and Fire (EIF) and Adaptive EIF, and the Spike Response Model which offers a “filtering” perspective of neuronal dynamics.

Neuromorphic Computing: Silicon Circuits for Common Neuron Models - This set of slides discusses and analyzes some silicon-based circuits for realizing a biological neuron. Included in the discussion are the simplest possible circuits such as a LIF neuron based on a leaky inverting integrator and the “axon hillock” circuit. These are followed by more advanced circuits such as the “sodium-potassium” circuit, the voltage-amplifier integrate and fire circuit, and the current mode conductance-based neuron circuit for the AdEx model.

Rate Encoding for Spiking Neural Networks

Temporal Time-to-First-Spike (TTFS) Encoding for Spiking Neural Networks 

Temporal Burst Encoding for Spiking Neural Networks 

Temporal Phase Encoding for Spiking Neural Networks - This set of four slides are individual treatments on different mechanisms of neuron encoding, specifically: (a) rate encoding, (b) time-to-first-spike encoding, (c) burst encoding, and (d) joint rate and phase encoding based on the phenomenon called “phase precession”.