Another incredible use of this tutorial is classifying 25 notes on a piano. Using a 450-50-26 model (I include a backgrounds sample), the results are mind-boggling!! In the list I show the note and how many times out of 331 samples it is predicted. My testing method is in the video. Next I need 1/2

@jmoenig.bsky.social I tried the MNIST data using a 10000 record csv. I used only 200 items to train in 784-100-10, and got through around 50 epochs before my computer hiccuped (likely memory?). It was enough to have what I consider to be a great start.

Here’s a video of sax versus clarinet. Even with poor playing it could tell the difference between these similar instruments.

Branford Marsalis quartet closing out Ottawa Jazzfest tonight. Accompanied by a giant dragonfly removing the pesky mosquitoes. It flies too fast to photograph.

It was Mavis Staples night tonight at Ottawa Jazzfest.

I’m looking forward to learning how to do this in Snap! as well. I made something similar in the past using @dalelane.co.uk Machine Learning for Kids, but it goes to Watson to create the models.

Apparently I need to say it is a video @deonandan.bsky.social

I forgot to tag people @sylviamartinez.bsky.social @garystager.bsky.social

So I can now share the video that goes with this post @jmoenig.bsky.social @kgiori.bsky.social