Skip to content

Commit

Permalink
Hai's articles
Browse files Browse the repository at this point in the history
  • Loading branch information
@garyericson
garyericson committed Dec 19, 2016
1 parent 0a711ee commit de2c4d3
Show file tree
Hide file tree
Showing 2 changed files with 39 additions and 26 deletions.
52 changes: 29 additions & 23 deletions articles/machine-learning/machine-learning-competition-faq.md
View file
Original file line number Diff line number Diff line change
Expand Up @@ -13,18 +13,18 @@ ms.workload: data-services
ms.tgt_pltfrm: na
ms.devlang: na
ms.topic: article
ms.date: 09/06/2016
ms.date: 12/16/2016
ms.author: haining;garye

---
# Microsoft Cortana Intelligence Competitions FAQ
**What is Cortana Intelligence Competitions?**

Microsoft is announcing Cortana Intelligence Competitions. Cortana Intelligence Competitions allows us to unite a global community of data enthusiasts by collectively solving some of world’s most complex data science problems. Cortana Intelligence Competitions allow data enthusiasts from across the world to compete and build highly accurate and intelligent data science models. Our hosted competitions are based on unique data sets that have been made available publically for the first time. Participants can win rewards or get recognition via our top 10 public leaderboard. Please go [here](http://aka.ms/CIComp) to access the Competitions home page.
The Microsoft Cortana Intelligence Competitions allow us to unite a global community of data enthusiasts by collectively solving some of the world’s most complex data science problems. Cortana Intelligence Competitions allow data enthusiasts from across the world to compete and build highly accurate and intelligent data science models. Our hosted competitions are based on unique data sets that have been made available publically for the first time. Participants can win rewards or get recognition via our top 10 public leaderboard. Please go [here](http://aka.ms/CIComp) to access the Competitions home page.

**How often will Microsoft release new competitions?**

We will be launching 1st party, Microsoft-owned competitions on a regular basis, approximately every 8-12 weeks.
We will be launching 1st-party, Microsoft-owned competitions on a regular basis, approximately every 8-12 weeks.

**Where can I ask general questions about data science?**

Expand All @@ -33,58 +33,64 @@ forum](https://social.msdn.microsoft.com/forums/azure/home?forum=MachineLearning

**How do I enter a competition?**

Access the Competitions home page via the Cortana Intelligence Gallery. This page contains all competitions that are running. Each competition will have detailed instructions and participation rules, prizes, and duration on its sign up page. Please go [here](http://aka.ms/CIComp) to access the Competitions home page.
Access the [Competitions](https://gallery.cortanaintelligence.com/competitions) home page in the [Cortana Intelligence Gallery](https://gallery.cortanaintelligence.com/), or go to [http://aka.ms/CIComp](http://aka.ms/CIComp). The home page lists all competitions that are currently running. Each competition has detailed instructions and participation rules, prizes, and duration on its sign-up page.

1. Find the competition you’d like to participate in Cortana Intelligence Gallery, read all the instructions and watch the tutorial video, then click on the “Enter Competition” button to copy the Starter Experiment into your existing Azure Machine Learning workspace. If you don’t already have access to a workspace, you must create one beforehand. Run the Starter Experiment, observe the performance metric, then use your creativity to improve the performance of the model. You will likely spend majority of your time in this step.
2. Create a Predictive Experiment with the trained model out of your Starter Experiment. Then carefully adjust the input and output schema of the web service to ensure they conform to the requirement specified in the Competition documentation. The tutorial document generally will have detailed instruction on how to accomplish this. You can also watch the tutorial video if available.
3. Deploy a web service out of your Predictive Experiment. Test your web service using the Test button or the Excel template automatically created for you to ensure it is working properly.
4. Submit your web service as the competition entry, and see your public score in the Cortana Intelligence Gallery competition page. And celebrate if you make into the leaderboard!
After you successfully submit an entry, you can go back to the copied Starter Experiment, iterate, and update your Predictive Experiment, update the web service, and submit an new entry.
1. Find the competition you’d like to participate in, read all the instructions and watch the tutorial video, then click the **Enter Competition** button to copy the Starter Experiment into your existing Azure Machine Learning workspace. If you don’t already have access to a workspace, you must create one beforehand. Run the Starter Experiment, observe the performance metric, then use your creativity to improve the performance of the model. You'll likely spend the majority of your time in this step.

2. Create a Predictive Experiment with the trained model out of your Starter Experiment. Then carefully adjust the input and output schema of the web service to ensure they conform to the requirements specified in the Competition documentation. The tutorial document generally will have detailed instructions for this. You can also watch the tutorial video, if available.

3. Deploy a web service out of your Predictive Experiment. Test your web service using the **Test** button or the Excel template automatically created for you to ensure it's working properly.

4. Submit your web service as the competition entry and see your public score in the Cortana Intelligence Gallery competition page. And celebrate if you make it onto the leaderboard!

After you successfully submit an entry, you can go back to your copied Starter Experiment, iterate, and update your Predictive Experiment, update the web service, and submit a new entry.

**Can I use open source tools for participating in these Competitions?**

The competition participants leverage Azure Machine Learning Studio, a cloud-based service within Cortana Intelligence Suite for development of the data science models and create Competition entries for submission. ML Studio not only provides a GUI interface for constructing machine learning experiments, it also allows you to bring your own R and/or Python scripts for native execution. The R and Python runtime in ML Studio come with a rich set of open source R/Python packages, and you can import your own packages as part of the experiment as well. ML Studio also has a built-in JuPyteR Notebook service for you to do free style data exploration. Of course, you can always download the datasets used in the Competition and explore it in your favorite tool outside of ML Studio.
The competition participants leverage Azure Machine Learning Studio, a cloud-based service within Cortana Intelligence Suite for development of the data science models and to create Competition entries for submission. Machine Learning Studio not only provides a GUI interface for constructing machine learning experiments, it also allows you to bring your own R and/or Python scripts for native execution. The R and Python runtimes in Studio come with a rich set of open source R/Python packages, and you can import your own packages as part of the experiment as well. Studio also has a built-in Jupyter Notebook service for you to do free style data exploration. Of course, you can always download the datasets used in the Competition and explore it in your favorite tool outside of Machine Learning Studio.

**Do I need to be a data scientist to enter?**

No. In fact, we encourage data enthusiasts, those curious about data science, and other aspiring data scientists to enter our contest. We have designed supporting documents to allow everyone to compete. Our target audience is:

* Data Developers, Data Scientists, BI and Analytics Professionals: those who are responsible for producing data and analytics content for others to consume.
* Data Stewards: those who have the knowledge about the data, what it means, and how it is intended to be used and for which purpose.
* Students & Researchers: those who are learning and gaining data related skills via academic programs in universities or participants of Massive Open Online Courses (MOOCs)
* **Data Developers**, **Data Scientists**, **BI** and **Analytics Professionals**: those who are responsible for producing data and analytics content for others to consume
* **Data Stewards**: those who have the knowledge about the data, what it means, and how it's intended to be used and for which purpose
* **Students** & **Researchers:** those who are learning and gaining data related skills via academic programs in universities, or participants in Massive Open Online Courses (MOOCs)

**Can I enter with my colleagues as a team?**

The Competition platform currently does not support team participation. Each competition entry is tied to a single user identity.
The Competition platform currently doesn't support team participation. Each competition entry is tied to a single user identity.

**Do I need to pay to participate in a competition?**

Competitions are free to participate in. You do, however, need access to an Azure Machine Learning workspace to participate. You can create a Free workspace without a credit card by simply logging in with a valid Microsoft account, or an Office 365 account. If you are already an Azure or Cortana Intelligence Suite customer, you can create and use a Standard workspace under the same Azure subscription. If you would like to purchase an Azure subscription you can go [here](https://azure.microsoft.com/pricing). Note the standard rates will apply when using a Standard workspace to construct experiments. See Azure Machine Learning pricing information [here](https://azure.microsoft.com/pricing/details/machine-learning/).
Competitions are free to participate in. You do, however, need access to an Azure Machine Learning workspace to participate. You can create a Free workspace without a credit card by simply logging in with a valid Microsoft account, or an Office 365 account. If you're already an Azure or Cortana Intelligence Suite customer, you can create and use a Standard workspace under the same Azure subscription. If you'd like to purchase an Azure subscription, go to the [Azure pricing](https://azure.microsoft.com/pricing) page. Note that the standard rates will apply when using a Standard workspace to construct experiments. See [Azure Machine Learning pricing information](https://azure.microsoft.com/pricing/details/machine-learning/) for more information.

[!INCLUDE [machine-learning-free-trial](../../includes/machine-learning-free-trial.md)]

**What are public and private scores?**

In most competitions, you will notice that you will receive a public score for every submission you make, generally within 10-20 minutes. But after the competition ends, you will receive a private score which is used for final ranking. Here is what happens:
In most competitions, you'll notice that you'll receive a public score for every submission you make, generally within 10-20 minutes. But after the competition ends, you'll receive a private score which is used for final ranking.

Here's what happens:

* The entire dataset used in the competition is randomly split with stratification into training and testing (the remaining) data. The random split is stratified to ensure that the distributions of labels in both training and testing data are consistent.
* The training data is uploaded and given to you as part of the Starter Experiment in the Import Data module configuration.
* The testing data is further split into public and private testing data, using the same stratification.
* The public testing data is used for the initial round of scoring. The result is referred to as public score and it is what you see in your submission history when you submit your entry. This score is calculated for every entry you submit. This public score is used to rank you on the public leaderboard.
* The private testing data is used for the final round of scoring after the Competition ended. This is referred to as private score.
* For each participant, a fixed number, that can vary depending on the competition, among your entries with the highest public scores are automatically selected to enter the private scoring round. The entry with the highest private score is then selected to enter the final ranking, which ultimately determined the prize winners.
* The public testing data is used for the initial round of scoring. The result is referred to as the public score and it's what you see in your submission history when you submit your entry. This score is calculated for every entry you submit. This public score is used to rank you on the public leaderboard.
* The private testing data is used for the final round of scoring after the Competition ends. This is referred to as private score.
* For each participant, a fixed number of entries with the highest public scores are automatically selected to enter the private scoring round (this number can vary depending on the competition). The entry with the highest private score is then selected to enter the final ranking, which ultimately determines the prize winners.

**Can Customers host a Competition on our platform?**

We welcome 3rd-party organizations to partner with us and host both public and private competitions on our platform. We have a competition onboarding team who will be happy to discuss the onboarding process for such competitions. Please get in touch with us at [[email protected]](mailto:[email protected]) for more details.

**What are the limitations for submissions?**

A typical competition may choose to limit the number of entries you can submit within a 24-hour span (UTC time 00:00:00 to 23:59:59), and the total number of entries you can submit over the duration of the competition. You will receive proper error messages when a limitation is exceeded.
A typical competition may choose to limit the number of entries you can submit within a 24-hour span (UTC time 00:00:00 to 23:59:59), and the total number of entries you can submit over the duration of the competition. You'll receive appropriate error messages when a limitation is exceeded.

**What happens if I win a Competition?**

Microsoft will verify the results of the private leaderboard, and then we will contact you. Please make sure that your email address in your user profile is up to date.
Microsoft will verify the results of the private leaderboard, and then we'll contact you. Please make sure that your email address in your user profile is up to date.

**How will I get the prize money if I win a Competition?**

Expand All @@ -96,11 +102,11 @@ The submission time is the tie-breaker. The entry submitted earlier outranks the

**Can I participate using Guest workspace?**

No. You must use a Free or a Standard workspace to participate. You can open the Competition starter experiment in a Guest workspace. But you will not be able to create a valid entry for submission.
No. You must use a Free or a Standard workspace to participate. You can open the Competition starter experiment in a Guest workspace, but you'll not be able to create a valid entry for submission from that workspace.

**Can I participate using a Workspace in any Azure region?**

Currently Competition platform only supports submitting entries from a workspace in South Central US Azure region. All Free workspaces live in South Central US. But if you choose to use a Standard workspace, please ensure it lives in South Central US Azure region. Otherwise your submission will not succeed.
Currently, the Competition platform only supports entries submitted from a workspace in the **South Central US** Azure region. All Free workspaces reside in South Central US, so you can submit an entry from any Free workspace. If you choose to use a Standard workspace, just ensure that it resides in the South Central US Azure region, otherwise your submission won't succeed.

**Do we keep Users’ Competitions Solutions/Entries?**

Expand Down
13 changes: 10 additions & 3 deletions articles/machine-learning/machine-learning-powershell-module.md
View file
Original file line number Diff line number Diff line change
Expand Up @@ -14,7 +14,7 @@ ms.workload: data-services
ms.tgt_pltfrm: na
ms.devlang: na
ms.topic: article
ms.date: 08/05/2016
ms.date: 12/19/2016
ms.author: garye;haining

---
Expand All @@ -23,6 +23,9 @@ The PowerShell module for Azure Machine Learning is a powerful tool that allows

You can view the documentation and download the module, along with the full source code, at [https://aka.ms/amlps](https://aka.ms/amlps).

> [!NOTE]
> The Azure Machine Learning PowerShell module is currently in preview mode. The module will continue to be improved and expanded during this preview period. Keep an eye on the [Cortana Intelligence and Machine Learning Blog](https://blogs.technet.microsoft.com/machinelearning/) for news and information.
## What is the Machine Learning PowerShell module?
The Machine Learning PowerShell module is a .NET-based DLL module that allows you to fully manage Azure Machine Learning workspaces, experiments, datasets, web services, and web service endpoints from Windows PowerShell.
Along with the module, you can download the full source code which includes a cleanly-separated [C# API layer](https://github.com/hning86/azuremlps/blob/master/code/AzureMLSDK.cs). This means you can reference this DLL from your own .NET project and manage Azure Machine Learning through .NET code. In addition, the DLL depends on underlying REST APIs that you can leverage directly from your favorite client.
Expand All @@ -47,8 +50,12 @@ Here's a quick example of using PowerShell to run an existing experiment:
For a more in-depth use case, see this article on using the PowerShell module to automate a very commonly-requested task: [Create many Machine Learning models and web service endpoints from one experiment using PowerShell](machine-learning-create-models-and-endpoints-with-powershell.md).

## How do I get started?
To get started with Machine Learning PowerShell, download the [release package](https://github.com/hning86/azuremlps/releases) from GitHub and follow the [instructions for installation](https://github.com/hning86/azuremlps/blob/master/README.md). You'll need to unblock the downloaded/unzipped DLL and then import it into your PowerShell environment. Most of the cmdlets require that you supply the workspace ID, the workspace authorization token, and the Azure region that the workspace is in. The simplest way to provide these is through a default config.json file, which is covered in detail in the installation instructions. Of course, you can also clone the git tree and modify/compile the code locally using Visual Studio.
To get started with Machine Learning PowerShell, download the [release package](https://github.com/hning86/azuremlps/releases) from GitHub and follow the [instructions for installation](https://github.com/hning86/azuremlps/blob/master/README.md). The instructions explain how to unblock the downloaded/unzipped DLL and then import it into your PowerShell environment.
Most of the cmdlets require that you supply the workspace ID, the workspace authorization token, and the Azure region that the workspace is in. The simplest way to provide these is through a default config.json file. The instructions also explain how to configure this file.

And if you want, you can clone the git tree, modify the code, and compile it locally using Visual Studio.

## Next steps
The PowerShell module will continue to be improved and expanded during this preview period. Keep an eye on the [Cortana Intelligence and Machine Learning Blog](https://blogs.technet.microsoft.com/machinelearning/) for more news and information.
You can find the full documentation for the PowerShell module at [https://aka.ms/amlps](https://aka.ms/amlps).

For an extended example of how to use the module in a real-world scenario, check out the in-depth use case, [Create many Machine Learning models and web service endpoints from one experiment using PowerShell](machine-learning-create-models-and-endpoints-with-powershell.md).

0 comments on commit de2c4d3

Please sign in to comment.