Azure ML is great. I would like Azure ML to support F# -> with both the code sample gen (currently C#) and the script support (currently R).

Would love to see a time series prediction model using a MultiClass Neural Net. Also, where's the N# documentation?

This is a critical package in calculating the relative impact of the components in a multivariate regression model.

Support C# Execution modules. While R and Python are useful in certain situations, C# is useful also. Given the large number of developers who know C# as opposed to R, this opens custom modules to a much wider audience.

Odbc drivers are not installed on machines running the python script. Solving this will enable users to create their own Reader module by using their own business logic to access azure sql.

We're using Stream analytics to write results in a line seperated json format, which is a standard for many other services, but it is impossible to read it from blob storage.
the workaround suggested will not help when using prediction experiment and getting the data from blob in a realtime manner

Currently only one storage account can be associated to an ML workspace, leaving SAS authentication as the only option to access other storage accounts. Nevertheless, testing the documented SAS access returns the comment:

[ModuleOutput] Azure Table with SAS Uri not implemented yet.

Implementing this feature, thus allowing access to multiple storage account, would widely increase the flexibility and applicability of ML to big data solutions.

A lot of the newer Algos require R or Python scripts to execute. LightGMB was built within Microsoft. It would be great if this could be added as a module within Studio ML rather than requiring R or Python.

https://github.com/Microsoft/LightGBM/issues

I get lots of these. A little bit more help finding out how to rectify them would be useful. Mainly on the training steps. I suspect its too many columns, but can't tell. Can more information be surfaced in the error logs?

In other machine learning tools, SVM supports different kernel functions, and each function may have additional parameters. I assume "Project to the unit-sphere" is a kernel function, but could not find a definition for that function.

A kernel I found useful in many cases was the Radial Basis Function (RBF, parameters C or lambda, gamma), but there are others. It would also be useful to allow defining a custom kernel function, for example with an R script (or whatever language is most appropriate for the SVM implementation used)

Would be easier to work with & compare multiple models. having to cross-pairing all models is too slow to design & hard to evaluate.

Even if the output when multiple models would have fewer features than the default 2 input evaluate. (ex: fewer visualization options, metrics) But at least a good subset of main evaluation metrics should be visible.

Many visual IDEs for Machine Learning are fast growing today.
They provide a way a mechanism to directly process data and enable consumption of data in popular BI tools. This should be done for Azure ML as well

Several custom algorithms I use are designed expecting input in JSON format. This is ideal for integrating with a web service, but unfortunately ML Azure doesn't seem to support it.

I would like to be able to connect the Application Insight Data from our WebApi directly into Machine Learning Experiment.
The goal for us to be able to predict things depending of the real time outcome

This is such a basic algorithm, I can't believe this is missing from ML

To avoid unnecessary resource consumption and get only those data needed from a table store please add a filter to the table store reader!

I want to expose an R package I wrote as a web service. Since it's a custom R package they docs say the install has to happen every time the user user my service.

There's going to be a lot of repetitive overhead going this route, Any chance I can pre-install my package so the user only has to consume using the web service?

With respect to Python in Azure ML, could opencv library access be provided? Currently, 'import cv2' fails. I am aware there are opencv modules and API in Azure, but I need to train over my own data which are not human faces.