General Availabilty Release of Azure Machine Learning!

Microsoft has announced the General Availability release of Azure Machine Learning. For details, please see this forum announcement:
https://social.msdn.microsoft.com/Forums/azure/en-US/home?forum=MachineLearning&announcementId=25a9893a-d3d8-4e4d-8a17-0f9bc60ec6a9

Hello Yordan,
Thanks for the quick answer.
Do you know if that's something in the plan? Any idea of a timeline you can share?

Similar Messages

**Free** 58-page e-book on using Azure Machine Learning and R

Free for download: "Data Science in the Cloud with Microsoft Azure Machine Learning and R" By Stephen Elston, Managing Director of Quantia Analytics, LLC.
To get you started creating and evaluating your own machine learning models, O’Reilly has commissioned a new report: “Data Science in the Cloud, with Azure Machine Learning and R.” We use an in-depth data science example — predicting bicycle rental demand —
to show you how to perform basic data science tasks, including data management, data transformation, machine learning, and model evaluation in the Microsoft Azure Machine Learning cloud environment. Using a free-tier Azure ML account, example R scripts, and
the data provided, the report provides hands-on experience with this practical data science example.
Specifically, this report shows you how to complete the following tasks using Azure ML and R:
◾Manage and transform data, using a highly scalable cloud environment
◾Build and evaluate machine learning models
◾Produce R graphics
◾Publish your models as web services
To download your free e-book, visit: https://azureinfo.microsoft.com/CO-Azure-CNTNT-FY15-02Feb-Data-Science-in-the-Cloud.html?ls=Media&lsd=Oreilly

Hi AK.
I'm running Windows 10 ETP with builtin IE, but the same problem is also with the latest Chrome 41. Using local DNS resolver with DNSSEC validation.
I'll try it later today at home and let you know.

Login problem with federated id to Azure Machine Learning

I log into Azure with a federated account ( organization account ). This works great for all the services I have used so far, but when I try to use the Azure Machine Learning product, it seems to require a Microsoft Live account.
Is there a way to log into Azure Machine Learning with a federated account?
I am using my personal account to post this since I could not log in with my federated account to fill in the form for the question.
Thanks

Sorry, at the moment we only support Live IDs. We are working on adding more support for authentication
Regards,
AK

Recommendations API by Azure Machine Learning - not working

I purchased the Recommendations API by Azure Machine Learning using the marketplace. When I go to the API explorer and the quickstart
screen, all i see is the rotating circles. I waited for about 10 minutes...no luck. I tried the sample app and it seem to build, run and work. Not sure what's going with the azure portal UI. Is
anyone really using it in prod?

This is a known issue in the Azure Management Portal, and it is being worked one. That said, it is only a visual glitch in the quick start page. You should be able to click on the DASHBOARD tab, and be able to access all the links required to use the Recommendations
API.
The marketplace in the management portal is in "PREVIEW" mode right now, and this is one of the issues that needs to be fixed there.
thanks!
-Luis Cabrera
Azure Machine Learning Marketplace.
Luis Cabrera. Program Manager -- Azure Machine Learning @luisito_cabrera Disclaimer: This post is provided &quot;AS IS&quot; with no warranties, and confer no rights.

Error using python multiprocessing in Azure Machine Learning

I want to do some parallel data manipulation in Azure ML but I can't seem to get multiprocessing to work.
So when I try this simple example in Azure ML Python Script Module
import multiprocessing
def azureml_main(dataframe1 = None, dataframe2 = None):
process_pool = multiprocessing.Pool(10)
l = []
for i in range(10):
l.append((0, 1000))
r = process_pool.map(do, l)
process_pool.close()
foo = 0
for a in r:
foo = foo + a
print(foo)
return dataframe1
def do(arg):
foo = 0
for i in range(arg[0], arg[1]):
foo = foo + 1
return foo
I get this error:
Error 0085: The following error occurred during script evaluation, please view the output log for more information:
---------- Start of error message from Python interpreter ----------
data:text/plain,Caught exception while executing function: Traceback (most recent call last):
File "C:\server\invokepy.py", line 167, in batch
odfs = mod.azureml_main(*idfs)
File "C:\temp\azuremod.py", line 19, in azureml_main
process_pool = multiprocessing.Pool(10)
File "C:\pyhome\lib\multiprocessing\__init__.py", line 232, in Pool
return Pool(processes, initializer, initargs, maxtasksperchild)
File "C:\pyhome\lib\multiprocessing\pool.py", line 159, in __init__
self._repopulate_pool()
File "C:\pyhome\lib\multiprocessing\pool.py", line 223, in _repopulate_pool
w.start()
File "C:\pyhome\lib\multiprocessing\process.py", line 130, in start
self._popen = Popen(self)
File "C:\pyhome\lib\multiprocessing\forking.py", line 280, in __init__
to_child.close()
IOError: [Errno 22] Invalid argument
---------- End of error message from Python interpreter ----------
When I try the same code in a local regular python file it works just fine
Any ideas? Or is there another way?

Hello Grahamaaa,
(a) Regarding map: Yes, vanilla map, list comprehensions, generators, etc. are all sequential.
(b) NumPy: You can call into multi-threaded native code (via OpenMP/TBB/etc.) as long as they follow the usual GIL locking/unlocking protocol. There are some calls such as SetCpuAffinity that are not supported.
There are essentially two ways to around the GIL in CPython: write C extensions that use threads or spawn multiple Python interpreters. The first approach should work modulo calls like SetCpuAffinity.
We are looking into why the fork call in Python is not working.
Regards,
Sudarshan (AzureML)

Azure Machine Learning : API for management (of experiments)

Hello there,
Not sure if I'm missing something, but I haven't found a way to create / alter an experiment via API and/or PowerShell commands (like we can define do via JSON/PowerShell in ADF).
I would also like to be able to publish / republish an experiment the same way to update a deployed service.
Thanks a lot ;)

Hello Yordan,
Thanks for the quick answer.
Do you know if that's something in the plan? Any idea of a timeline you can share?

Work with reader in Machine learning

hi,
i am doing work on azure machine learning, and i found one doubt, suppose i have 100 records right now and i create experiment. it gives me some result.
now suppose after 2-3 days i have 2000 records so do i have to run the experiment again to use large scale of data or azure machine learning experiment will get those newly added records when i call web service using RRS.
Please provide some guidance.

When you train the service on the new 2000 records the model will be re-trained from scratch. That is, the knowledge obtained from the old 100 records will be lost.
The feature you're talking about is called
incremental/online learning, which is something we don't support yet.
-Y-

Microsoft Azure for machine learning on data collected from microsoft spectrum observatory

Dear Sir,
Can any one please suggest how shall I apply Microsoft Azure for applying machine learning techniques on spectrum data available from Microsoft spectrum observatory . Please send a mail.
Regards,
pavan

Hi Pavan!
Apologies, I'm not familiar with MSO - how are you collecting the data provided there? Is there an API? Do you have access to the raw data?
Regards,
AK

Machine Learning Text Analytics Service - Maximum String Length

In the documentation here: http://azure.microsoft.com/en-us/documentation/articles/machine-learning-apps-text-analytics/
it doesn't provide a max length value for the string parameters. In other words, what's the largest string I can upload to the API?
Thanks,
ATOMBOMB

As far as HTTP protocol can support it. For e.g. if you are using OData endpoint with get, then max will be determined by URI length.

Performance of Machine Learning Webservice API

Hi Team,
I have some questions about the performance of Machine Learning Web Service API.
I've recently grabbed the tweeter sentiment analysis model sample and tested the web service performance in both single request/response and batch execution mode. Based on the result, single request API takes about 5 seconds to process one
tweet, which is not applicable to real-time processing as the speed that tweet get generated is much faster than this. So I have to buffer tweet stream and send batch execution requests which takes about 0.25 sec per tweet when sending 500 tweets per batch.
The questions are:
1. Given a task such as tweeter sentiment analysis, can the API running under single request mode do better than a few seconds per record, in order to enable real-time processing? Currently I am using the free version ML service.
Could standard version have better performance in terms of processing single web service request?
2. How to evaluate the performance and estimate the execution time/pricing of a model built on Azure ML? From the pricing web page it says "The best way to estimate prediction API call times is to benchmark a model on the Machine Learning
service.". Can anyone explain this a bit more in details about how to "benchmark a model" ?
Thanks

Hi Dexter,
It is likely that the initial calls to the web service are triggering model initialization and are taking longer than steady state calls. To measure the performance of a fully initialized service - you would need to send multiple concurrent requests at once.
There is an unofficial tool that helps you do this today:
https://github.com/paveldournov/RRSWarmer but you can do that with simple client code as well.
There is an option in the ML APIs to send multiple entries (tweets in your case) in a single request to the Request/Response service. In this case - all of the entries will be sent to the model at once, which can be faster per each tweet, while it will not
require asynchronous batch processing. Sending 500 entries per call in this case should not be a problem.
The free workspace allows you to build services that can scale only to 4 concurrent calls at a time. A paid workspace will allow scaling this to 20 concurrent calls by default, and to 200 with an API call to adjust the max concurrent calls. The performance
of an individual call should not change with going from Free to Paid, but the amount of call volume and rps is up to 50 times more in the Paid case.
Re. the pricing - the resources on this page are outlining the key pricing points:
http://azure.microsoft.com/en-us/pricing/details/machine-learning/ . The API pricing is composed out of two components - the number of calls and the time to process the calls.
I'd recommend to setup a web service and do perf testing of the service from a single client sending serialized requests. This will allow you to measure the compute time per call. Then a simple multiplication by the expected number of calls should give
an idea about the total cost.
Thanks!

External LDAP port in Azure machine

Hi all,
I have two machines. One is used as an Active Directory domain controller machine. One is for web server. I want to capture LDAP queries by network package analysis tool but don't know if Azure machine uses different port rather than standard port 389. If
that's the external port so how can I know that port?
Your suggestion would be greatly appreciated.
Regards,
-T.s
Thuan Soldier
A 23-year-old man loving Microsoft technologies and making crazy ideas on business journey.
SharePoint Vietnam |
Blog | Twitter

Hello Thuan Soldier,
The port 389 can be used to request information from the local DC. However, the requesting application can obtain all of the attributes for those objects. For example, a request to port 389 could be used to obtain a user’s department
The port 3268 is used for queries specifically targeted for the global catalog. LDAP requests sent to port 3268 can be used to search for objects in the entire forest. However, only the attributes marked for replication to the global catalog can be returned.
For example, a user’s department could not be returned using port 3268 since this attribute is not replicated to the global catalog.
Thanks,
Syed Irfan Hussain

Is there any one who knows how to apply machine learning algorithms to the spectrum data collected using ni usrp for ism band to use spectrum efficiently by link adaptation

Hi,
may i know any one working in cognitive radio research and applying machine learning techniques using the spectrum data collected from NI usrp kits

Can't edit my message anymore, so for the tldr crowd (too long didn't read), here is a shorter version:
The good:
E4200 worked fine for about 15 hours with great throughput and link quality (with one brief disconnect in the middle - online gaming software shows brief disconnects that otherwise go totally unnoticed).
The bad:
Wireless on the E4200 stopped broadcasting entirely (to a Windows 7 laptop with a 802.11n 2.4 GHZ USB adapter).
I had switched between two adapters that use the same chipset, same driver and in fact show up as a single device in device manager) without rebooting the router but it worked (fine) for an hour after I swapped the adapters.
Right before I rebooted it, I checked and the E4200 was not hot to the touch and wired internet was still working after the wireless radio stopped working.
A reboot of the router cured it.
The open question as to the root cause:
Now I want to know whether this wireless radio ceasing to broadcast (requires a reboot of the E4200) will happen daily?
I.e. I want to know whether it it really is overheating (then again why would only some have that problem needing to reboot daily while others have gone weeks with the E4200 without a single reboot being needed and without any problems occurring) or whether it had to do with changing the adapters back and forth without rebooting the router. Keep in mind that the router worked fine for an hour or so after I stopped swapping the adapters.
Extra question:
Will getting a Cisco AE1000 USB adapter perhaps help? (I know this is a loaded question and their is no easy / sure answer, but even a "maybe" or reasons why it might help at this point would be better than nothing.
Any similar experiences (with the router needing to be rebooted to get wireless radio back on) would be appreciated as it may help myself and others experiencing these types of issues.

The Azure team has released the Azure Diagnostics extension for Linux VMs

You can now ingest Syslog data into OMS from Azure Linux VMsThe Azure team has released the Azure Diagnostics extension for Linux VMsIf you use Azure Diagnostics, Operations Management Suite can ingest syslog data from storage accounts configured to store your diagnostics data. As of the publication of this blog post, OMS can ingest all logs routed through syslog or RsyslogD.
To enable this functionality, follow the steps below:1. Enable Azure Diagnostics for your Linux machines2. In the Classic Azure Portal find your Operational Insights workspace.Note: If you can’t find your OMS workspace here, it means you haven’t linked your workspace to your Azure Subscription – follow the steps to link workspaces here
This topic first appeared in the Spiceworks Community

The scripts you need for Linux will be posted on OTN soon.

Access denied error while running opening remote power-shell session from one azure machine to another

We are trying to open a remote power-shell session from one azure machine to another azure machine and that is failing with the following error. Any ideas?
Observations: -
If the worker role is running under network service or local system, then it fails.
We emulated the same behavior from a stand-alone exe and that also failed when the exe was run under network service or local system.
If we run the same stand-alone exe from a user account, then it works.
Unhandled Exception: System.Management.Automation.Remoting.PSRemotingTransportException: Connecting to remote server sampleapprm.cloudapp.net failed with the following error message : Access is denied. For more information, see the about_Remote_Troubleshooting
Help topic.
   at System.Management.Automation.Runspaces.AsyncResult.EndInvoke()
   at System.Management.Automation.Runspaces.Internal.RunspacePoolInternal.EndOp
en(IAsyncResult asyncResult)
   at System.Management.Automation.Runspaces.Internal.RemoteRunspacePoolInternal
.Open()
   at System.Management.Automation.RemoteRunspace.Open()
   at ConsoleApplication2.Program.Main(String[] args)

Hi,
This error only show access is denied, and it is hard for me to deal with this issue, may be Windows Powershell forum or Windows Azure VM are good forums for you, could you give more details information to us for a better help if this issue
was related with windows Azure.
Best Regards
We are trying to better understand customer views on social support experience, so your participation in this interview project would be greatly appreciated if you have time. Thanks for helping make community forums a great place.
Click
HERE to participate the survey.

Test Run - L1 and L2 Regularization for Machine Learning

James McCaffrey looks at how regularization can help to reduce overfitting, a common result of machine learning that undermines the predictive accuracy of the model.
Read the article

Hi,
Jaya and I discussed this and there are a couple of issues that may be the problem in your case:
Filter selection might be selecting different features in the training and scoring experiment. Try removing filter based feature selection module.
Feature hashing preserves the original text features causing the input dimension to increase considerably. Try removing the original text column using "Project Columns" module (after "Feature Hashing" and prior to "Train")
Limit the size of the model (there's a limitation on operationalizable models) by restricting hashing bits (say 10-12 bits)
Here's a screen shot of scoring experiments (where I restricted to 10 bits)
regards
Sharat

General Availabilty Release of Azure Machine Learning!

Similar Messages

Maybe you are looking for