Machine Learning

Welcome! You can use this site to tell the Azure Machine Learning team what features you would like to see.
Azure Machine Learning enables developers and data scientists to build advanced analytic web services in minutes and hours, eliminating the heavy lifting associated with deploying machine learning in modern data driven applications.

How can we improve Microsoft Azure Machine Learning?

You've used all your votes and won't be able to post a new idea, but you can still search and comment on existing ideas.

There are two ways to get more votes:

  • When an admin closes an idea you've voted on, you'll get your votes back from that idea.
  • You can remove your votes from an open idea you support.
  • To see ideas you have already voted on, select the "My feedback" filter and select "My open ideas".
(thinking…)

Enter your idea and we'll search to see if someone has already suggested it.

If a similar idea already exists, you can support and comment on it.

If it doesn't exist, you can post your idea so others can support it.

Enter your idea and we'll search to see if someone has already suggested it.

  1. 13 votes
    Vote
    Sign in
    (thinking…)
    Password icon
    Signed in as (Sign out)
    You have left! (?) (thinking…)
    0 comments  ·  Flag idea as inappropriate…  ·  Admin →
  2. Enable GPU support when deploying to Azure Container Instances

    Azure Container Instances can be configured to provide GPU resources to the container (https://docs.microsoft.com/en-us/azure/container-instances/container-instances-gpu). However, there is no way to ask for GPU resources when deploying ML services via the Azure ML SDK (AciWebservice.deploy_configuration) or Azure CLI (az ml service create aci). Even if the image is built with GPU support enabled (ContainerImage.image_configuration(enable_gpu=True,...)), the image does not work properly in ACI because the GPU resources are not present.

    Please extend the SDK and CLI to allow specifying the GPU count and SKU when deploying to ACI.

    13 votes
    Vote
    Sign in
    (thinking…)
    Password icon
    Signed in as (Sign out)
    You have left! (?) (thinking…)
    1 comment  ·  Flag idea as inappropriate…  ·  Admin →
  3. Reference Web Service Parameters in R/Python/SQL/SQLite/etc scripts

    Currently, individual operation parameters can be "Set as web service parameter". However, there are many cases where I don't want to pass the whole value as a parameter, but reference a parameter within that value. This happens mostly in script operations (R, Python, SQL Reader, Apply SQL Operation).

    For example, I may have a SQL query with various filters. A R script may also want to access various parameters to alter its behavior. This can be useful when experimenting with various model options, or to generalize (reuse) an experiment for different cases.

    This could be implemented by allowing, in all…

    13 votes
    Vote
    Sign in
    (thinking…)
    Password icon
    Signed in as (Sign out)
    You have left! (?) (thinking…)
    1 comment  ·  Flag idea as inappropriate…  ·  Admin →
  4. Add Power Query datasets published to the Corporate Data Catalog as a source in Azure Machine Learning

    It would be fantastic if we could use Power Query datasets saved to the Corporate Data Catalog as a source in Azure Machine Learning.

    Here is a blog post in which I describe it in more detail and outline the benefits. http://www.christoolivier.com/missed-opportunity-power-query-in-azure-machine-learning/

    12 votes
    Vote
    Sign in
    (thinking…)
    Password icon
    Signed in as (Sign out)
    You have left! (?) (thinking…)
    0 comments  ·  Flag idea as inappropriate…  ·  Admin →
  5. 12 votes
    Vote
    Sign in
    (thinking…)
    Password icon
    Signed in as (Sign out)
    You have left! (?) (thinking…)
    0 comments  ·  Flag idea as inappropriate…  ·  Admin →
  6. Add [snap to] Grid Options

    Add a grid view allowing to lock blocks at grid position.
    Improves view and readability.

    11 votes
    Vote
    Sign in
    (thinking…)
    Password icon
    Signed in as (Sign out)
    You have left! (?) (thinking…)
    0 comments  ·  Flag idea as inappropriate…  ·  Admin →
  7. R Execute - Create an Output Zip Pin

    Allow the user to save a list of R objects (dataframes, models, functions, vectors, ...) to an .Rdata zip file.

    This would allow the user to build an R model and save it (along with any other R object or function) for use in other R execute blocks.

    11 votes
    Vote
    Sign in
    (thinking…)
    Password icon
    Signed in as (Sign out)
    You have left! (?) (thinking…)
    0 comments  ·  Flag idea as inappropriate…  ·  Admin →
  8. notebooks.azure.com could be the ideal interface for: AzureML, HDInsight Spark + Jupyter & Batch

    It would be great to have one notebook interface that could connect to all the relevant services.

    This would also allow us to access notebooks, while a HDInsight cluster is stopped.

    11 votes
    Vote
    Sign in
    (thinking…)
    Password icon
    Signed in as (Sign out)
    You have left! (?) (thinking…)
    1 comment  ·  Flag idea as inappropriate…  ·  Admin →
  9. Add Github integration to azure machine learning for custom code/packages/repository

    Please make it possible to link a custom repository on github (with custom python code) to a dataset. This makes usage much more convenient instead of uploading a new zip everytime code changes.

    11 votes
    Vote
    Sign in
    (thinking…)
    Password icon
    Signed in as (Sign out)
    You have left! (?) (thinking…)
    0 comments  ·  Flag idea as inappropriate…  ·  Admin →
  10. Fix the input and output names in AML Studio Web Service

    AML Studio gives the ability to change the input and output names for a Web Service. But changing these values does not have any effect, and the published Web Service always uses the default values input1, output1, etc... These values are used in the JSON schema of the Web Service, and it would be nice to be able to change them.

    Apparently Microsoft is already aware of this (see https://social.msdn.microsoft.com/Forums/azure/en-US/03e5f824-abad-4e19-b57d-fd3bfc5c7f1e/ml-web-service-inputoutput-names-reverting-to-default). Please fix this issue.

    11 votes
    Vote
    Sign in
    (thinking…)
    Password icon
    Signed in as (Sign out)
    You have left! (?) (thinking…)
    1 comment  ·  Flag idea as inappropriate…  ·  Admin →
  11. Workspace ARM resource should repair missing permissions on other Azure objects (e.g. Key Vault)

    The Microsoft.MachineLearningServices/workspaces ARM resource should be able to fix missing permissions on Azure resources it depends on (such as Key Vault access policy) when it is redeployed.

    Currently, the ML Workspace creates a AAD Service Principal for itself and assigns it at least these permissions (and perhaps more) during provisioning:
    - Contributor access to the Resource Group
    - Contributor access to the Container Registry
    - an access policy in the Key Vault allowing all operations except Purge
    - Storage Blob Data Contributor access to the Storage Account

    However, if anything happens to these permissions (for example, the Key Vault access…

    11 votes
    Vote
    Sign in
    (thinking…)
    Password icon
    Signed in as (Sign out)
    You have left! (?) (thinking…)
    0 comments  ·  Flag idea as inappropriate…  ·  Admin →
  12. Export Python code from AzureML

    Exporting Python script from an experiment setup in AzureML. Would be a great feature to have.

    11 votes
    Vote
    Sign in
    (thinking…)
    Password icon
    Signed in as (Sign out)
    You have left! (?) (thinking…)
    0 comments  ·  Flag idea as inappropriate…  ·  Admin →
  13. Add possibility to extend timeout time for web service request

    I'm trying to build a web service that does much more than just taking some already processed data and runs it through a model to determine some classification. This service should be able to take raw data, process to specified format, do some NLP computing on the data using trained databases and return the result to the user. When raw data is larger that around 200kb, the service does not have the time to finish before getting a timeout, and I have no ide why it cannot be extended more than 90s. It could be solved using batch request, but…

    11 votes
    Vote
    Sign in
    (thinking…)
    Password icon
    Signed in as (Sign out)
    You have left! (?) (thinking…)
    planned  ·  0 comments  ·  Flag idea as inappropriate…  ·  Admin →
  14. Add a 2nd output port to Feature Hashing Module

    The Feature Hashing module would be improvedif it had a 2nd output port that output the data set of the hash mapping (ie the hash value and original string pairs). This would aide in debugging, troubleshooting, and tuning since the mappings and collisions could be visualized and analyzed.

    11 votes
    Vote
    Sign in
    (thinking…)
    Password icon
    Signed in as (Sign out)
    You have left! (?) (thinking…)
    0 comments  ·  Flag idea as inappropriate…  ·  Admin →
  15. Output Multiple Result Sets in R Execute Script Module

    Here is a common use case:

    library(caret)
    trainIndex <- createDataPartition(df$serviceName, p = .8, list = F, times = 2)
    df.train <- df[trainIndex,] # I need to export this, the training data
    df.test <- df[-trainIndex,] # and, I need to export this, the testing data

    The createDataPartition is useful becuase it ensures both training and testing data have the same factors.

    11 votes
    Vote
    Sign in
    (thinking…)
    Password icon
    Signed in as (Sign out)
    You have left! (?) (thinking…)
    1 comment  ·  Flag idea as inappropriate…  ·  Admin →
  16. Would like support for BUGS (Bayesian inference using Gibbs Sampling)

    Would like support for WinBUGS or OpenBUGS
    (Bayesian inference using Gibbs Sampling) program in AzureML.

    10 votes
    Vote
    Sign in
    (thinking…)
    Password icon
    Signed in as (Sign out)
    You have left! (?) (thinking…)
    0 comments  ·  Flag idea as inappropriate…  ·  Admin →
  17. Integrate with Visualization Tool(s)

    Visualizations need to be included somewhere that seamlessly works with Azure ML, and preferable also integrated with Data Factory. This way you will have a true end to end data solution.

    Being able to create dashboards like the new Azure Portal, with many different types of visualizations would amazing. Have not looked into Power BI close enough to know whether this will provide the type of visualizations that predictive models require. Is Microsoft working on an Azure Visualisations?

    9 votes
    Vote
    Sign in
    (thinking…)
    Password icon
    Signed in as (Sign out)
    You have left! (?) (thinking…)
    0 comments  ·  Flag idea as inappropriate…  ·  Admin →
  18. Access to logs of Execute R Script in Web Service

    Currently, there is an option to enable logging which can be turned on in the web service configuration page. Unfortunately, those logs are kind of useless to me.

    What I would like to get is all the logs from Execute R Script modules available in Azure Storage after each run with logging set to on. Exactly the same ones I can access after running an experiment, containing also text from the console.

    Without proper logs I can't easily debug the script. You mentioned over a year ago that you were going to include debugging which would also be very helpful.

    9 votes
    Vote
    Sign in
    (thinking…)
    Password icon
    Signed in as (Sign out)
    You have left! (?) (thinking…)
    0 comments  ·  Flag idea as inappropriate…  ·  Admin →
  19. Partial Dependence plots

    Similar to R, Azure should create a module of partial dependence plots (PDP) for its native machine learning modules.
    PDP, wherein, the predictor variable of interest is varied over its range while the remaining predictors are fixed at a joint set of values sampled from the dataset to produce an instance of the response dependence curve. This process is replicated for all learn records, sampling new joint set of fixed values each time, creating a family of partial dependence curves which are then averaged and centered so as to generate the final partial dependence plot (PDP).

    9 votes
    Vote
    Sign in
    (thinking…)
    Password icon
    Signed in as (Sign out)
    You have left! (?) (thinking…)
    0 comments  ·  Flag idea as inappropriate…  ·  Admin →
  20. Import PMML directly (without external call to R or Python)

    Provide module which reads / imports PMML directly (without external call to R or Python)

    9 votes
    Vote
    Sign in
    (thinking…)
    Password icon
    Signed in as (Sign out)
    You have left! (?) (thinking…)
    1 comment  ·  Flag idea as inappropriate…  ·  Admin →
  • Don't see your idea?

Machine Learning

Feedback and Knowledge Base