Machine Learning
Welcome! The Azure Machine Learning team invites you to share and vote for features to help you build, manage, and deploy custom machine learning models.
Have a technical question, or want to learn more? Please visit our documentation, MSDN forum or StackOverflow.
-
Local (software) version of Azure ML
A version of Azure ML to download and install on your local machine under a specific price to buy it. You will be able to use your local machine power and most import, load your files without the need to upload them. Then some kind of connection between the offline and online version to sync them in case you want to work on another machine or on the online version of Azure ML.
20 votesdeclined ·
Adminmluservoice
(Program Manager, Microsoft Azure Machine Learning, Microsoft Azure)
responded
Azure ML is part of Azure offering and will continue to be offered in cloud
-
Make the .dlls available in Visual Studio
I am working with Azure ML for about a week now and I think it is awesome. Great job. I do notice that the amount of spaghetti on the designer grows pretty quickly. Can you explose the .dlls (like Microsoft.Analytics.Modules.EvaluateGeneric.Dll) in Nuget and build some Visual Studio designer support? That way I could write the same stuff in F# on my machine. Not really Azure, but it is Microsoft ML...
12 votesdeclined ·Adminmluservoice
(Program Manager, Microsoft Azure Machine Learning, Microsoft Azure)
responded
Unfortunately, this request is not part of the roadmap for the next 12 months.
-
Status bar for time left on an experiment or % complete
Some big data experiments can be quite large. Sometimes it is hard to tell if the experiment is still running or got caught in a never ending loop. Having some indicator to let me know how a experiment is progressing would be nice.
11 votesdeclined ·Adminmluservoice
(Program Manager, Microsoft Azure Machine Learning, Microsoft Azure)
responded
Unfortunately, this request is not part of the roadmap for the next 12 months.
-
9 votesdeclined ·Adminmluservoice
(Program Manager, Microsoft Azure Machine Learning, Microsoft Azure)
responded
Currently, LSTM is supported in CNTK and in future roadmap we will integrate Azure ML with CNTK to provide this capability
-
Octave Support
For those of us that prefer Octave over R or have some code samples in Octave, support of Octave is a good feature in AzureML.
7 votesdeclined ·Adminmluservoice
(Program Manager, Microsoft Azure Machine Learning, Microsoft Azure)
responded
We don’t have direct plan to support Octave as execution environment within Azure ML but we are working on enabling support for arbitrary code along with bring your own compute that will allow user to bring their own choice of libraries to execute within Azure ML
-
Image classification
After the recently added Image Reader it is now possible to create a image classification model. However the model is not possible to publish as a web service due to the huge amount of parameters (for me appr. 180K).
The best solution to this would be to make it possible to enter a image into the webservice, convert it via Image Reader and then make the classification. Another, less appealing, solution would be to make it possible to enter a huge amount of parameters into the web service.4 votesdeclined ·Adminmluservoice
(Program Manager, Microsoft Azure Machine Learning, Microsoft Azure)
responded
Unfortunately, we do not have plans to support that large of a number of parameters at this time. We are separately releasing consumable APIs (available today in the Gallery) which allow you to upload an image and then classify it several ways.
-
MATLAB Support (via MATLAB Compiler)
For those of us that prefer matlab to R and have licenses that we can bring in for compiler, adding MATLAB Support (via MATLAB Compiler) would let us use the same model that we develop locally to run/exec in production instead of having to re-code in R or plugin to the AzureML models, which reduces the test/debug/production cycle, and gives the same output in dev/test as in production.
4 votesdeclined ·Adminmluservoice
(Program Manager, Microsoft Azure Machine Learning, Microsoft Azure)
responded
Thanks for your suggestion, but unfortunately we are not planning for this feature in in our roadmap at this time.
-
AzureML Studio API
We are considering using AzureML and to do so we'd like to build our own management UI around it to do things that are pertinent to our business, for example we want to give our data scientists the opportunity to nominate a particular experiment as the "production" version so that consuming systems know which experiement API to call.
To do this we would need AzureML itself to have an API that would enable us to (e.g.) ask for a list of a user's experiments. The returned info would tell us the API endpoint for those experiments.
Stated another way perhaps,…
3 votesdeclined ·Adminmluservoice
(Program Manager, Microsoft Azure Machine Learning, Microsoft Azure)
responded
Thank you for the suggestion. Unfortunately, we do not plan to provide this functionality in the near future.
-
prediction
How do you make a prediction from API?
3 votesdeclined ·Adminmluservoice
(Program Manager, Microsoft Azure Machine Learning, Microsoft Azure)
responded
Please see the following documentation: https://azure.microsoft.com/en-us/documentation/articles/machine-learning-consume-web-services/.
In the future, please post product questions to the forum at https://social.msdn.microsoft.com/forums/azure/en-US/home?forum=MachineLearning
-
Export model to a C# project so that it can be further extended programmatically.
The idea is that C# code is generated from the model definition. This code could be edited, compiled and deployed to Azure.
This would enable powerful scenarios and it would be possible to overcome many of the limitations in ML today. It would also allow developers to create custom modules for transformations and use those in their code.
3 votes -
Support State Space Model
State Space Model is convinient for prediction PV.
3 votesdeclined ·Adminmluservoice
(Program Manager, Microsoft Azure Machine Learning, Microsoft Azure)
responded
This feature is not part of the roadmap for near future.
-
Input: Reference Data Refresh Cycle(Scheduled..)
Providing functionality to refresh the Reference Data.
3 votesdeclined ·Adminmluservoice
(Program Manager, Microsoft Azure Machine Learning, Microsoft Azure)
responded
I believe this is covered via our re-training capabilities. Please see: http://blogs.technet.com/b/machinelearning/archive/2015/03/03/programmatic-retraining-of-azure-machine-learning-models.aspx
-
Show free tier experiments in Azure Portal
I have signed up for the free tier of Azure ML at http://studio.azureml.com. I think it's fantastic. However, I also have an Azure account and it would be great to access my ML experiments within the Portal with my other Azure resources.
3 votesdeclined ·Adminmluservoice
(Program Manager, Microsoft Azure Machine Learning, Microsoft Azure)
responded
The free tier experiments will not show up in your Azure subscription as it is not tied to a ML Workspace in your subscription. Please feel free to create a ML Workspace in your subscription and then you will be able to copy your experiment from your free subscription to your standard user account.
-
Allow work on the model while it is running.
I have been waiting 30 minutes for my SQL query to resolve. It would be nice if I could go ahead and add more modules to my workspace while this is happening. Obviously nothing should hookup to the reader that is running, but it would be nice to drop more modules on to the canvas and hook them up.
2 votesdeclined ·Adminmluservoice
(Program Manager, Microsoft Azure Machine Learning, Microsoft Azure)
responded
Unfortunately, this request is not part of the roadmap for the next 12 months.
-
WebService User Login Capabilities in Azure ML.
Can we have webservice user login capabilities in Azure ML i.e. who has accessed the webservice and when.
1 votedeclined ·Adminmluservoice
(Program Manager, Microsoft Azure Machine Learning, Microsoft Azure)
responded
Unfortunately, this is not part of our planning at this time. Perhaps we may reconsider next year.
-
Cardiovascular monitoring
Predict long term and near real time cardiovascular disease based on ISS driven monitoring solution using wearable devices coupled via Bluetooth and NFC to Windows hone and Windows tablet apps.
1 votedeclined ·Adminmluservoice
(Program Manager, Microsoft Azure Machine Learning, Microsoft Azure)
responded
This particular solution is not in our roadmap to be developed by our team at the moment. However, the users of the community are always adding new experiments so please check back often!
-
Spare parts
Estimate the amount of products of components will be returned in the future.
1 vote -
regions and subscriptions not showing when creating new ML
drop downs only contain one item
1 votedeclined ·Adminmluservoice
(Program Manager, Microsoft Azure Machine Learning, Microsoft Azure)
responded
By design. Only one region is currently supported. As more regions become available, more values will appear in the dropdown.
-
Model Description Language
It would be great if models or workflows in AzureML can be encoded in XML or JSON format (I call it Model Description Language). This way sharing models, documenting models, working on different versions of the same model would be easier than the current approach which is sharing models via email. It would be much like WSDL + WS-BPEL. In addition, documentation of steps and input/output of each step can be done automatically and it can be easier to reproduce the model in desktop environment (like RStudio) as well as design surface of AzureML.
1 votedeclined ·Adminmluservoice
(Program Manager, Microsoft Azure Machine Learning, Microsoft Azure)
responded
Thank you for your suggestion; however, this request is currently not part of the roadmap for the next 12 months.
-
Train Model console output
Please provide console output for "train model" during development. Right now, its a blackbox, once we hit Run in ML studio, simply wait for it to complete.. a console output showing % complete or some sort of estimation is essential..
1 votedeclined ·Adminmluservoice
(Program Manager, Microsoft Azure Machine Learning, Microsoft Azure)
responded
Thank you for your suggestion; however, this request is currently not part of the roadmap for the next 6 months.
- Don't see your idea?