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.
-
"compare to" in visualization
I am inrolled in the online course "essentials of data science" and in many of the videos we can go to a model, visualize it and show scatterplots/diagrams by chosing "compare to". When I try this myself with my own models there are, in some of the cases, no "compare to". It is a really useful function when working with feature engineering, why is it only available in some cases? How can I make it available all the time?
24 votes -
Make azure ML studio scalable
For large datasets and complex algorithms it either takes very long to train the model or training fails because of memory lack. Please make azure ml studio scalable. Let users choose cores and memory before running an experiment
24 votes -
Support Decimal / Money Datatypes
When importing from Azure SQL Warehouse, support decimal and money datatypes (or autocast to float)
23 votes -
Visualization graphs - Exportable
I am trying to use the 'compare' feature in the visualization window, where it shows you graphs that compare two columns, but it is hard to see. The window isn't resizable and the axis-labels are too small/cut short so I can't actually see what it is I am analyzing. It would be perfect if you could export this graph -- even if to a JPEG or PDF, so you could visualize it better. Thank you!
23 votesplanned ·
Adminmluservoice
(Program Manager, Microsoft Azure Machine Learning, Microsoft Azure)
responded
Thank you; the team will review as part of our planning process.
-
23 votes
-
Add filter capability to reader
To avoid unnecessary resource consumption and get only those data needed from a table store please add a filter to the table store reader!
22 votes -
About character code of Machine Learning Workbench
Currently there is no Shift-JIS in the encodable character code.
I want Shift-JIS to be supported as well.＜In Japanese＞
現在はエンコーディング可能な文字コードにShift-JISがありません。
Shift-JISもサポートされるようにしてほしい。22 votes -
Please fix this error package 'gridExtra' is not available (for R version 3.1.1) to let me run the course
Please fix this error package 'gridExtra' is not available (for R version 3.1.1) to let me run the course
22 votes -
Add email notifications when a running instance completes
If you've got a big dataset, often the first few steps (e.g. Reader getting several Gb of data from Azure BLOB storage, or a split/fold/partition/etc on a big dataset) can take a disproportionately long time. And that's fine, but I don't necessarily want to sit & watch.
If you added an "email me when step X completes" so I can go make dinner while the import runs, I would be a happy, happy boy.
22 votesunder review ·Adminmluservoice
(Program Manager, Microsoft Azure Machine Learning, Microsoft Azure)
responded
Thanks. We will look into it.
-
Additional input datasets for R / Python scripts
In order to have an R or Python script use more than two datasets I have to create many modules and do the combinations incrementally. It would make the UI much easier to work with if I could connect an indeterminate number of datasets as inputs.
A potentially acceptable solution would be to include the ability to create a dataset of datasets prior to using as an input to a script module.
22 votes -
Integrate the Nupic HTM CLA
Integrate the Nupic HTM CLA with the Machine Learning platform. I'm not seeing an obvious way to utilise both Nupic (which I really like) and Azure ML which looks excellent.
21 votesunder review ·Adminmluservoice
(Program Manager, Microsoft Azure Machine Learning, Microsoft Azure)
responded
Nupic is a deep learning algorithm that Numenta is developing and promoting. The closest thing that is currently part of AML is building a multi-layered, neural network, but there are lots of fundamental differences between NuPIC and vanilla neural networks. It will take a substantial amount of engineering for it to be incorporated, and we will evaluate if it will fit in the backlog.
-
Re-open feature request "Deployment of ML experiments programmatically"
This feature request is marked as complete https://feedback.azure.com/forums/257792-machine-learning/suggestions/7575534-deployment-of-ml-experiments-programmatically, but the library is still far from being complete.
In https://github.com/hning86/azuremlps#import-amlexperimentgraph, we see the following note:
"Please note that the exported JSON file only contains references to the exact instance and version of the assets (modules, trained models, datasets, etc.). The assets themselves are NOT serialized into the JSON file. As a consequence, when you import it back into the Workspace, unless you are importing it back into the same Workspace, or unless your graph only contains global assets such as built-in modules, you will not be able to create…
20 votes -
Add feature selection methods using classification/regression algorithms
Just for an example, I want to select features with classification/regression algorithms such as Multi Layer Perceptron, Linear regression, Decision tree and so on, selecting the best feature subset by performing forward/backward Best-first search.
This is can be done in WEKA.
20 votes -
Allowing sorting of values from Visualize screen
Values should be sortable from the Visualize screen by clicking on the column and sorting from ascending to descending or vice versa.
20 votes -
Allow for data split into training, validation and testing using only one split
Allow for data split into training, validation and testing using only one split.
Coming from a SAS background I find the split module to be very clunky and the use two split modules in order to split the data into 3 sets (training, validation, testing) to be very counter-productive.
19 votesunder review ·Adminmluservoice
(Program Manager, Microsoft Azure Machine Learning, Microsoft Azure)
responded
We are considering adding this.
-
Import/Export experiments and all related assets in a file
I would like to be able to have a functionality within the Studio for importing exporting the experiments in a file (possibly JSON format but even a proprietary binary would be fine as well).
The file should contain all the assets (not only their references like it happens now) so that I can seamlessly move experiments from different environments (QUAL-PROD etc.) and different subscriptions.
Moreover this would allow to better share the experiments within the community storing the file for e.g. in github.
19 votes -
Standardize canvas navigation
Management Studio diagrams:
Scroll = Scroll
Ctrl + Scroll = ZoomSSIS canvas, SSAS diagrams in Visual Studio:
Scroll = Scroll
Ctrl + Scroll = ZoomML Studio:
Scroll = ZoomYes, ML Studio is browser-based but navigating across the canvas could be similar to previous products.
I don't have the numbers to prove the following statement - but I think that vertical scrolling is used more often than zooming.19 votesunder review ·Adminmluservoice
(Program Manager, Microsoft Azure Machine Learning, Microsoft Azure)
responded
Thank you; the team will review as part of our planning process.
-
18 votesplanned ·Adminmluservoice
(Program Manager, Microsoft Azure Machine Learning, Microsoft Azure)
responded
This feature is under review.
-
Desktop client for AML Studio
The new AML Services offering includes a desktop client, AML Workbench. Please create a desktop client for AML Studio as well. Personally, I find the Web UI for AML Studio a little clumsy sometimes. Ability to use local datasets in AML Studio desktop client would also be nice.
Ideally, AML Studio could be just another feature of the AML Workbench client. No reason to make matters too complicated.
18 votes -
Get serious with Datasets! Beat Kaggle
If you are serious about ML Studio, you'd rather include hundreds and hundreds sample datasets. We should see 2-3 samples for each algorithm. Two sets for each dataset needed: 1. real, dirty data 2. the same but clean data. So user could concentrate on cleaning or on the algorithm side of the problem. Datasets should be in categories and with tags... you have fantazy, right?
Samples for different user categories: students, scientists, developers, kids.
Samples for different tasks: study, competitions, science (Yes! Please, include datasets for *real* problems in science)
We should beat Kaggle in this aspect (why not ImageNet?)17 votes
- Don't see your idea?