MNIST, MNIST, MNIST…. but Spiking?
Using Spiking Convolutional Neural Networks on the “Hello World” of AI
Ok, I know you’ve heard of MNIST before. It’s basically the “Hello World” of AI, but this article’s different!
And if you don’t know what MNIST is: it’s a dataset of about 70,000 labeled images of hand-drawn numbers which is literally perfect for training AI classification algorithms. That’s why MNIST is the benchmark dataset for all sorts of AI algorithms and structures, including the one that I’ll be writing about today.
The third generation of Artificial Neural Networks
They’re called Spiking Neural Networks, or SNNs for short.
SNNs are a type of neural network based off the brain. Just like the brain’s neurons communicate by spiking, SNNs communicate through spike trains.
Spike trains are sequences of discrete, binary values that are based in time. The technical words are asynchronous and sparse which means that the spikes don’t all happen at the same time and that they don’t occur that often. This mechanism is unlike normal normal neural networks which output continuous…