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.


  1. Is it possible to change the web service URI from ".../score" to a custom name?

    I'm trying to replace an existing web service in Microsoft Dynamics that was using a model from Azure Machine Learning Studio. It has a hardcoded endpoint suffix like "/predict". For the Azure ML Service I'm building to replace it, there seems to be a hardcoded suffix called "/score" instead in the web service URI.

    Can I customize this? I'd like to replace "/score" with "/predict" etc so that I can swap out the new model.

    1 vote
    Vote
    Sign in
    (thinking…)
    Sign in with: Microsoft
    Signed in as (Sign out)
    You have left! (?) (thinking…)
    0 comments  ·  Flag idea as inappropriate…  ·  Admin →
  2. Improve model portability

    Allow easy identification of all of an ML studio models components and exporting of these to new workspaces. Currently copying has major limitations such as data set size, workspace location etc. Datasets get duplicated when used in multiple models during the copy etc. It should be possible to select and copy multiple experiments or projects and have the copy process bring along everything that is needed including training experiments, predictive experiments, transformations and web service parameters etc. without creating duplication.

    1 vote
    Vote
    Sign in
    (thinking…)
    Sign in with: Microsoft
    Signed in as (Sign out)
    You have left! (?) (thinking…)
    0 comments  ·  Flag idea as inappropriate…  ·  Admin →

    Dear Azure Customer,

    Thanks for the feedback. Your feedback is very important to us. Our engineering team will investigate it and consider your feedback seriously. We will get back to you soon.

    Thank you for understanding.
    Azure CXP Community

  3. Mark models as faulty

    We would like to manually mark some models as faulty so that they cannot be used or restored.

    1 vote
    Vote
    Sign in
    (thinking…)
    Sign in with: Microsoft
    Signed in as (Sign out)
    You have left! (?) (thinking…)
    1 comment  ·  Flag idea as inappropriate…  ·  Admin →
  4. Create notebook from docker image

    I'd like to create a notebook from a docker image in my private container registry so that I can test out the pieces of my azure ml experiment that failed. It seems like you can't create a notebook with a docker image or conda environment which makes it tedious to set up a custom environment. Thanks for considering this feature!

    1 vote
    Vote
    Sign in
    (thinking…)
    Sign in with: Microsoft
    Signed in as (Sign out)
    You have left! (?) (thinking…)
    triaged  ·  1 comment  ·  Flag idea as inappropriate…  ·  Admin →
  5. AML pipeline with ADF step

    Is there a plan to support ADF as an AML pipeline steps ?
    The use case would be to leverage the extended ADF pipeline capabilities to prepare Data before it is used for training.
    Also, Data Engineer and Data Scientist would benefit from Data Lineage understanding through feature like Data Flow.
    I appreciate your feedbacks.

    1 vote
    Vote
    Sign in
    (thinking…)
    Sign in with: Microsoft
    Signed in as (Sign out)
    You have left! (?) (thinking…)
    0 comments  ·  Flag idea as inappropriate…  ·  Admin →
  6. Is there a way to attach Data Bricks Cluster which is in a different resource group

    Is there a way to attach Data Bricks Cluster which is in a different resource group

    1 vote
    Vote
    Sign in
    (thinking…)
    Sign in with: Microsoft
    Signed in as (Sign out)
    You have left! (?) (thinking…)
    need-feedback  ·  2 comments  ·  Flag idea as inappropriate…  ·  Admin →
  7. We need to build in algorithms for Multi-label classification. Currently we do not support any of BP-MLL or ML-kNN etc.

    There are no support for Multi label classification by default in ML services. Users have to write their own code in Python/R or any language and go ahead with an ensemble of models. Why dont we have any support by default?

    1 vote
    Vote
    Sign in
    (thinking…)
    Sign in with: Microsoft
    Signed in as (Sign out)
    You have left! (?) (thinking…)
    0 comments  ·  Flag idea as inappropriate…  ·  Admin →
  8. ML Services: View registered datastores and datasets in the workspace GUI blade

    The ability to be able to register datasets and datastores is really powerful. Currently, if I register datasets / datastores and their metadata with the SDK, I can't find them in the workspace GUI in the portal (where I see my experiments / runs / compute / registered models etc) - I have to list them programmatically.

    It would be great if there were a tab added to the workspace portal blade so that I could see all the datasets / datastores, the metadata I have added to them, their version history etc.

    1 vote
    Vote
    Sign in
    (thinking…)
    Sign in with: Microsoft
    Signed in as (Sign out)
    You have left! (?) (thinking…)
    1 comment  ·  Flag idea as inappropriate…  ·  Admin →
  9. Make warning that optimiation hasn't converged more apparent.

    The evaluation of my Multiclass Neural Network wasn't all that good. In looking to see what might have gone wrong, I came across a curious warning message in the training logs:
    /azureml-envs/azureml1f1665f56e65fc1b774765297713b3cf/lib/python3.6/site-packages/sklearn/neuralnetwork/multilayer_perceptron.py:562: ConvergenceWarning: Stochastic Optimizer: Maximum iterations (100) reached and the optimization hasn't converged yet.

    It would be good if this warning was more obvious. I'm re-running now with >100 iterations, but wouldn't have known to do this if I didn't comb through the logs.

    1 vote
    Vote
    Sign in
    (thinking…)
    Sign in with: Microsoft
    Signed in as (Sign out)
    You have left! (?) (thinking…)
    0 comments  ·  Flag idea as inappropriate…  ·  Admin →
  10. Remove AKS minimum core restriction ("cluster_purpose")

    If I want to attach an AKS cluster in development environments that don't meet the 4 agents / 12 core requirements, I need to set the "clusterpurpose = AksCompute.ClusterPurpose.DEVTEST" parameter via the Python API. However, this parameter is not supported in ARM templates and also not in the Azure CLI extension, so this complicates our automation strategy.

    The minimum requirements for running an AKS cluster in production are well documented, so it shouldn't be up to the "Machine Learning service" to enforce any restrictions.

    Having this "cluster_purpose" parameter is just an unnecessary complication, and should therefore be removed…

    1 vote
    Vote
    Sign in
    (thinking…)
    Sign in with: Microsoft
    Signed in as (Sign out)
    You have left! (?) (thinking…)
    0 comments  ·  Flag idea as inappropriate…  ·  Admin →
  11. Support Parquet file format as Import Data action

    Support for Parquet data format as Import Data, as many big data projects are using this format. It is much lighter then CSV. This format is already supported by Azure Data Factory.

    1 vote
    Vote
    Sign in
    (thinking…)
    Sign in with: Microsoft
    Signed in as (Sign out)
    You have left! (?) (thinking…)
    0 comments  ·  Flag idea as inappropriate…  ·  Admin →

    Hi Azure customer,

    Thank you for the feedback here, could you please share more details about what product you are mentioning? Is that Azure machine learning studio, Visual Interface, or you mean for all products ^^?

    Waiting for your response. Thank you very much.

    Regards,
    Azure CXP Community

  12. Show summary in run history

    Show the summary in run history so that you can more easily track your iterations in your experiment

    1 vote
    Vote
    Sign in
    (thinking…)
    Sign in with: Microsoft
    Signed in as (Sign out)
    You have left! (?) (thinking…)
    0 comments  ·  Flag idea as inappropriate…  ·  Admin →

    Dear Azure Customer,

    Thanks for the feedback. Your feedback is very important to us. Our engineering team will investigate it and consider your feedback seriously. We will get back to you soon.

    Thank you for understanding.
    Azure CXP Community

  13. When are you planning to upgrade the CRAN R version on ML. Is this an option or not on the cards for the near future?

    Is this on your road map or not? Need to know for purposes of functional technology

    1 vote
    Vote
    Sign in
    (thinking…)
    Sign in with: Microsoft
    Signed in as (Sign out)
    You have left! (?) (thinking…)
    0 comments  ·  Flag idea as inappropriate…  ·  Admin →
  14. Add sklearn Pipeline for feature engineering chaining

    When doing repeated cross validation like k-fold, some feature transformer like normaliser/maxminscaler/feature_selector, etc should be fitted only on the training data (and then tranform the training and validation set). Otherwise some data leakage occurs.

    Therefore, would it be possible to add the sklearn pipeline object as an argument ?
    After splitting the dataset into training and validation set, automl would fit on the training set and apply the pipeline on the validation set, whatever the cv strategy

    One could then add the whole dataset and use the available cross validations strategies without

    1 vote
    Vote
    Sign in
    (thinking…)
    Sign in with: Microsoft
    Signed in as (Sign out)
    You have left! (?) (thinking…)
    0 comments  ·  Flag idea as inappropriate…  ·  Admin →
  15. Add the Kolmogorov-Smirnov test within the "Filter Based Feature Selection" module

    The Kolmogorov-Smirnov test provides very important metrics for assessing the information value in each column, see https://en.wikipedia.org/wiki/Kolmogorov%E2%80%93Smirnovtest#Two-sampleKolmogorov–Smirnov_test

    If you are interested, I can provide an Excel spreadsheet with relevant example.

    1 vote
    Vote
    Sign in
    (thinking…)
    Sign in with: Microsoft
    Signed in as (Sign out)
    You have left! (?) (thinking…)
    0 comments  ·  Flag idea as inappropriate…  ·  Admin →
  16. Easier authentication into Azure ML ACR with Azure CLI

    Currently authentication requires several steps:

    amlacrid=$(
    az ml workspace show -g $rg -w $ws \
    --query containerRegistry -o tsv
    )

    amlacr=$(
    az resource show --id "$aml
    acr_id" \
    --query name -o tsv
    )

    az acr login -n "$aml_acr”

    1 vote
    Vote
    Sign in
    (thinking…)
    Sign in with: Microsoft
    Signed in as (Sign out)
    You have left! (?) (thinking…)
    0 comments  ·  Flag idea as inappropriate…  ·  Admin →
  17. Pivot and unpivot needed come on this is common datascience component

    Pivot and unpivot needed come on this is common datascience component

    1 vote
    Vote
    Sign in
    (thinking…)
    Sign in with: Microsoft
    Signed in as (Sign out)
    You have left! (?) (thinking…)
    triaged  ·  1 comment  ·  Flag idea as inappropriate…  ·  Admin →
  18. ML studio Experiments page to allow folders

    In ML Studio, at experiments, we need to allow a user to create folders to organize various projects. Usually one creates different experiments to tackle the problem in different ways. In the end, maybe only one or two experiments are put to use. However, the user would like to retain all prior experiments for future reference on what was tried. It also helps to allow user to add a Readme.txt to go along with each project to document. In our case, a workspace is shared, and the person who created the experiments may leave. In addition, after two years I…

    1 vote
    Vote
    Sign in
    (thinking…)
    Sign in with: Microsoft
    Signed in as (Sign out)
    You have left! (?) (thinking…)
    1 comment  ·  Flag idea as inappropriate…  ·  Admin →
  19. Integrate with Azure Time Series Insights

    Azure ML allows me to collect inputs going into the prediction service. Why not integrate with the awesome and powerful Azure Time Series Insights service to help me analyze how this data is changing over time?

    1 vote
    Vote
    Sign in
    (thinking…)
    Sign in with: Microsoft
    Signed in as (Sign out)
    You have left! (?) (thinking…)
    0 comments  ·  Flag idea as inappropriate…  ·  Admin →
  20. 1 vote
    Vote
    Sign in
    (thinking…)
    Sign in with: Microsoft
    Signed in as (Sign out)
    You have left! (?) (thinking…)
    0 comments  ·  Flag idea as inappropriate…  ·  Admin →
  • Don't see your idea?

Machine Learning

Categories

Feedback and Knowledge Base