RE: ACE - Blog #4 - The neural network

jduchniewicz — Fri, 14 May 2021 14:16:56 GMT

Great post and a lot of useful information [View:/resized-image/__size/16x16/__key/commentfiles/f7d226abd59f475c9d224a79e3f0ec07-4c8ed2bb-071d-44ef-bb79-2d13db409f27/contentimage_5F00_1.png:16:16] It seems like e14 could compile a whole ML tutorial section from contributions to this project [View:/resized-image/__size/16x16/__key/commentfiles/f7d226abd59f475c9d224a79e3f0ec07-4c8ed2bb-071d-44ef-bb79-2d13db409f27/contentimage_5F00_2516.png:16:16]!

What you could also do, is to utilize K-Fold Crossvalidation. This way you can create more resilient models and if your dataset is not big enough it can help significantly. It reduces bias in the model, showing how it fares against unseen data.
If you want an example, I used it in my ML model training.

RE: ACE - Blog #4 - The neural network

dubbie — Fri, 14 May 2021 11:41:37 GMT

A very good Blog, and a very good description, as well as some good results.

Dubbie

RE: ACE - Blog #4 - The neural network

DAB — Wed, 12 May 2021 19:48:43 GMT

Nice update.

The biggest problem I see is getting a good set of "fall" data sets that can reliably alert the system to a fall.

DAB