Let’s talk convolutional layers.
They're everywhere in ML, but proving them in ZK sounds scary.
Turns out… it’s faster to validate a CNN conv layer than run it. 🤯
How? Reduce it to polynomial math + apply sumcheck + FFT tricks.
Math wins again.
They're everywhere in ML, but proving them in ZK sounds scary.
Turns out… it’s faster to validate a CNN conv layer than run it. 🤯
How? Reduce it to polynomial math + apply sumcheck + FFT tricks.
Math wins again.