Neural Network Predictive Tool -- Deep Learning with Convolution and Pooling

[JohnJPS]

This request is largely based on the implementation found on AzureML; (take their free trial and check out the Deep Convolutional and Pooling NN example from their gallery).  This allows you to specify custom convolutional and pooling layers in a deep neural network. This is an extremely powerful machine learning technique that could be tricky to implement, but could perhaps be (for example) a great initial macro wrapped around something in Python, where currently these are more easily implemented than in R.

[Atabarezz]

I second the availability of deepNN capabilities in Alteryx

though without being able to utilise a GPU these computation will take so long hours

eventually using alteryx will become a torture...

[BrianCalvanese]

I think this could be built using nested macros... would be really cool to run this in-DB on Snowflake!

[JohnJPS]

KNIME (open source) has done some interesting thinking on Guided Analytics.

In particular they have an entire example (requiring enterprise version to run, so no longer open source) of Automated Machine Learning that looks not unlike like Uber's Ludwig, except it ready-to-go as point-and-click from a browser, including ample room for user input. Very robust. Would love to see something like that in Alteryx. Seems very doable.

[SyxRyx]

I think it's a must have... 

Not much you can do with a single layer NN...

Just curious as to why Alteryx could not be using GPU for these computations, when available on the user's machine, like the NVIDIA GPU? Agree it's not going to be ultra fast, but better than CPU.

[JohnJPS]

Given the support for both R and Python, and the fact that both of these now support tensorflow and possibly other awesome frameworks, this idea can probably be put to rest.  It would be neat to have a point/lick interface to add/modify layers in a network (e.g. which would auto-generate Keras-ish code behind the scenes), but that would be a tall order.

[SyxRyx]

The whole idea of Alteryx is to not have to write code in R or Python... Like with all the other predictive models already available (Random Forest, Regressions, Decision Trees, ...).

Maybe it's too "difficult" to implement in Alteryx, but saying you can use R or Python within Alteryx is not a solution, in my opinion. It defeats the purpose of the platform.

[Atabarezz]

@SyxRyx great point and I agree, a flexible neural network tool with interactive grahics is a need

If I'm able to solve it with R and Python there will be no need for licensing Alteryx for 5k usd

[JohnJPS]

This is a pretty old request of mine... given python integration, a neat macro would be as good as anything.