Jack Dermody

Glad to hear about Linux. Agree that it would be great to use OpenCL. However, it's not just a dependency on CUDA unfortunately. There's also cublas and cusolve (for SVD)...

No, there's no lower level API unfortunately. But you could adapt your data, set the batchsize to N and train for 1 iteration to get what you want... You would...

You would need to create a custom Softmax activation to achieve that. You could use the existing softmax activation as a guide but instead of calculating softmax over the entire...

Example code for https://www.nuget.org/packages/BrightWire/ (v3) is here: https://github.com/jdermody/brightwire/tree/master/ExampleCode Example code for https://www.nuget.org/packages/BrightWire.Net4/ (v2) is here: https://github.com/jdermody/brightwire-v2/tree/master/SampleCode

I see your point - the current design is that everything is created through indirection and the classes themselves are not public from the assembly. The drawback of this is...

Hi, sorry it took so long to get back to you! I think you want to create a table with an input vector column and a target output vector column....

I'll take a look at the IndexOutOfRangeException but the correct nuget package should be: `Install-Package BrightData.Numerics` This has been corrected in the BrightWire github readme

There's no spatial or sequential aspect to a shopping list so a classic feed forward neural network is your best bet. ML techniques other than neural networks will work for...

Sorry, I don't have any time at the moment. However, you could try mapping the date with sin or cos into a cyclic feature for your model.