Details
Poster
Presenter(s)
![Nasrollahi Seyed Amirhossein Headshot](https://confcats-catavault.s3.amazonaws.com/CATAVault/ieeecass/master/files/styles/cc_user_photo/s3/user-pictures/72071.jpg?h=8a338780&itok=yDy9MSti)
Display Name
Nasrollahi Seyed Amirhossein
- Affiliation
-
AffiliationConcordia University
- Country
Abstract
In a neural network, the first layer interfaces between the external world and the remainder of the network. We propose using the well-known Δ∑ encoder as the basis in the design of two voltage-driven input-layer spiking neuron circuits. They convert analog voltages into spike trains with firing rates linearly proportional to the input voltage. We use available circuits: a 1st order Δ∑ modulator, D-flipflops, a differential-pair synapse, and an Integrate-and-Fire (IF) neuron. These input-layer neurons can be implemented on the same IC as the rest of the SNN and can encode values over a wide range of inputs.