Core Concepts
Lu.i is a parametrizable electronic implementation of the leaky-integrate-and-fire neuron model designed for educational use and scientific outreach, enabling visualization and hands-on experience of neuronal dynamics and spiking neural networks.
Abstract
The paper introduces Lu.i, a low-cost electronic neuron designed for educational and outreach purposes in neuroscience. Lu.i implements the leaky-integrate-and-fire (LIF) neuron model, which captures the fundamental properties of neuronal information processing.
The key highlights of the Lu.i system are:
Hardware implementation: Lu.i is a printed circuit board (PCB) that physically realizes the LIF neuron dynamics through analog electronic circuits. This allows for a tangible, hands-on experience of neuronal behavior.
Configurable parameters: Lu.i offers control over various neuron and synapse parameters, such as time constants, synaptic weights, and polarity. This enables exploring the impact of these parameters on neuronal computation.
Visualization and interfacing: Lu.i features on-board LEDs to visualize the membrane potential and spike output, allowing standalone operation. It also provides interfaces for external equipment like oscilloscopes and microcontrollers for more advanced experiments.
Low-cost and accessible design: The PCB has been optimized for cost-effective manufacturing, with a unit price around $3 even for small batches. This makes Lu.i accessible for educational institutions and outreach activities.
The paper demonstrates several experiments that can be conducted with Lu.i, ranging from illustrating the basic LIF neuron dynamics to building small spiking neural networks that perform simple computational tasks like logic gates. Lu.i has been actively used in workshops, classrooms, and science communication events to nurture the understanding of neuroscience research and neuromorphic engineering among students and the general public.
Stats
The membrane potential Vmem(t) of the LIF neuron model is governed by the differential equation:
Cmem dVmem(t)/dt = -gleak [Vmem(t) - Vleak] + Isyn(t)
The synaptic current Ij_syn(t) for a presynaptic spike j at time tj_pre follows an exponential kernel:
Ij_syn(t) = wi * exp(-(t - tj_pre)/τsyn)
The refractory period of the neuron is approximately 12 ms.
Quotes
"Lu.i features current-based synaptic inputs that enable the formation of simple spiking neural networks (SNNs) and offers control over many parameters, including the time constants and the synaptic weights."
"Lu.i was designed to illustrate two of the fundamental aspects of biological neurons: spatio-temporal accumulation of input and event-based communication, both of which are captured by the LIF model."
"Lu.i complements a range of pedagogical tools spanning from experimental to computational neuroscience, combining the advantages of both approaches."