BigData, Analytics and Machine learning are new buzzwords everywhere you look and its easy to get overwhelmed with information when you decide to start ramping up on these! As part of an internal experiment we are doing at Cennest we’ve logged out our “learning path” for Machine learning which would take an application developer to at least a beginner level.
Below is our learning path to beginner level for Azure Machine Learning
Step 1: – Create an Azure account
Step 2:- See all the videos at http://azure.microsoft.com/en-us/trial/get-started-machine-learning-b/ //Notice the -b here? They seem to be doing A/B testing and well we liked –b 
Step 3:- Next the following two videos/articles from the Documentation. Needless to say keep trying the same stuff in your own Azure ML Studio!
http://azure.microsoft.com/en-us/documentation/videos/getting-started-with-ml-studio/
http://azure.microsoft.com/en-us/documentation/articles/machine-learning-create-experiment/
Step 4: – Complete the end-to-end example here. Its a bit tedious but follow through and create this experiment later blindfolded (OK may have gotten a bit extreme here…)
Step 5: – Absolutely loved this beginner tutorial by Aditi Technologies. We had planned to do one just like this at the end of this learning path but they totally beat us to it:-) . In-fact we will be using the same Azure ML data source they’ve used i.e Adult Census data set from the Azure Data Samples for our next blog!
http://blog.aditi.com/cloud/windows-azure-machine-learning-beginners-sample/
Step 6: – This one explains the basic steps very nicely
http://bluewatersql.wordpress.com/2014/08/01/azure-machine-learning-a-deeper-look/
Few Key points for beginners: -
- You can get sample data to play with from http://archive.ics.uci.edu/ml/
- Use Project Column filters from Data Transformation filters to filter out unwanted columns.
- Use Missing Values Scrubber from Data Transformation filters to filter out null values/Rows from your data
- Once you have created an experiment. Save a copy of it before you create a Trained Model out of it. This will help you edit the experiment again. You cannot regenerate the experiment from the trained model if you want to edit it later (at least at the time of this writing!)
- Once the Experiment is published as a web service you should test it out using the “Test” link provided in the Dashboard.
- If you want to now access the Web service from outside the console then you can make use of the code provided by Azure in the API Help page. Note that the .NET code is valid for a Console application
We have a few tricks up our sleeve wrt easing the pain of testing your own ML Experiment. Watch this space for more!!
Update:- Now you can access your published ML Experiment via our online console at http://mlonazure.cloudapp.net/ . Read more about it here.
Drop us a note if you like the learning path and encourage us to capture more of these as we get into other technologies soon!!
Till then!
Team Cennest