Skip to main content

Azure Machine Learning (Day 13 of 31)

List of all posts from this series: http://vunvulearadu.blogspot.ro/2014/11/azure-blog-post-marathon-is-ready-to.html

Short Description 
Machine Learning is a new service time on Microsoft Azure that can be used with success for predictive analytics. It can be used with success for machine learning AS A SERVICE. You don’t need to make any kind of hardware configuration, system configuration and so on. Everything is offered as a service, you only need to define the source, what algorithms and mechanism you want to define and this is all.

Main Features 
Easy configuration
Defining input/output sources, what kind of algorithms model you want to define and so on can be made very easily.
Configuration as a Workflow
The configuration portal is simple and very simple to use. You have the ability to configure everything from portal, using drag and drop very similar with a workflow. Because of this is easy to learn and use. For people is very natural.
Algorithm
There is a long list of predefined algorithms that can be used by default. This algorithms can be used as they are with success. Below you can find some algorithms that are available:

  • Scalable Boosted Decision trees
  • Bayesian Recommendation Systems
  • Deep Neural Networks 
  • Decision Jungles

R Language
You have the ability to write your own algorithm in language R and import in Machine Learning. In this way you can use Machine Learning for any kind of use cases.
Fast deploy and analyzed
The configuration and deployment is extremely fast and easy. In a few minutes you can configure it and start it. In this way you can have more focus on your data and how you can configure it to get the expected output. For focus in machine ‘training’ and less on system configuration.
No hardware and software configuration
Yes, you don’t need to make any kind of configuration. You only use it as it is.    
Training models
You have the ability to train and test you models. There is full support to create and train your own models.
Monitoring
Once you created your model, it is simple to monitor your model, see how it work and check the performance of it.
Data Scientist 
It is oriented to data scientist that can focus on model, experiments and scoring. For them and machine learning world this is more important than everything else.
REST API
Yes, each configuration is available over RESTfull.
Batch execution service 
There is full support for batches. You can take data from Azure Storage, HD Insight, SQL Azure or HTTP resources. It is very simple to ingest data from different sources.
Scoring support
You can check for machine learning alg. Based on the generated score and check how good is your solution.
Publish as a Service
Once you have your solution validate, you can publish it as a service simple. In a few clicks you can get your own model publish as a service that can be used by external sources. You can even sell it as a service.

Limitations 
I have only a little experience on Machine Learning. Because of this I’m not in the position to give feedback related to this topic.

Applicable Use Cases 
Personally I would use it in the fallowing use cases:
Real time logs analytics
I would use Machine Learning to analyze logs in real time to detect strange behavior and trigger different alerts to our support team.
Device Health Status
If you have a lot of devices and you want to detect the health of device and predict when they will fail, than Machine Learning is an interesting tool for that.
Possible Security Problems
Machine Learning can be used with success to check what users are doing on the system and check if the behavior of them is not the one that is expected and what is the normal one (based on history).

Code Sample 
# Map 1-based optional input ports to variables
mydata <- maml.mapInputPort(1) # class: data.frame

data.split <- strsplit(mydata[1,1], ",")[[1]]
data.split <- sapply(data.split, strsplit, ";", simplify = TRUE)
data.split <- sapply(data.split, strsplit, ";", simplify = TRUE)
data.split <- as.data.frame(t(data.split))

data.split <- data.matrix(data.split)
data.split <- data.frame(data.split)

# K-Means Cluster Analysis
fit <- kmeans(data.split, mydata$k) # k-cluster solution

# get cluster means 
aggregate(data.split,by=list(fit$cluster),FUN=mean)
# append cluster assignment
mydatafinal <- data.frame(t(fit$cluster))
n_col=ncol(mydatafinal)
colnames(mydatafinal) <- paste("V",1:n_col,sep="")

# Select data.frame to be sent to the output Dataset port
maml.mapOutputPort("mydatafinal");
Code sample source: http://azure.microsoft.com/en-us/documentation/articles/machine-learning-r-csharp-cluster-model/

Pros and Cons 
Pros
  • Easy to use
  • Expose as a service very easily
  • R support
  • Scoring support
  • Multiple ingest sources supported
  • Salable
Cons
-

Pricing
When you calculate the cost of Machine Learning you should take into account:

  • How long the system runs
  • How many predictions are done
  • Outbound traffic
Conclusion
In conclusion I would say that Machine Learning is it an interesting service that should be checked by data scientist and people that are working with big data. There is an business opportunity there that is very interesting.

Comments

Popular posts from this blog

Windows Docker Containers can make WIN32 API calls, use COM and ASP.NET WebForms

After the last post , I received two interesting questions related to Docker and Windows. People were interested if we do Win32 API calls from a Docker container and if there is support for COM. WIN32 Support To test calls to WIN32 API, let’s try to populate SYSTEM_INFO class. [StructLayout(LayoutKind.Sequential)] public struct SYSTEM_INFO { public uint dwOemId; public uint dwPageSize; public uint lpMinimumApplicationAddress; public uint lpMaximumApplicationAddress; public uint dwActiveProcessorMask; public uint dwNumberOfProcessors; public uint dwProcessorType; public uint dwAllocationGranularity; public uint dwProcessorLevel; public uint dwProcessorRevision; } ... [DllImport("kernel32")] static extern void GetSystemInfo(ref SYSTEM_INFO pSI); ... SYSTEM_INFO pSI = new SYSTEM_INFO(

Azure AD and AWS Cognito side-by-side

In the last few weeks, I was involved in multiple opportunities on Microsoft Azure and Amazon, where we had to analyse AWS Cognito, Azure AD and other solutions that are available on the market. I decided to consolidate in one post all features and differences that I identified for both of them that we should need to take into account. Take into account that Azure AD is an identity and access management services well integrated with Microsoft stack. In comparison, AWS Cognito is just a user sign-up, sign-in and access control and nothing more. The focus is not on the main features, is more on small things that can make a difference when you want to decide where we want to store and manage our users.  This information might be useful in the future when we need to decide where we want to keep and manage our users.  Feature Azure AD (B2C, B2C) AWS Cognito Access token lifetime Default 1h – the value is configurable 1h – cannot be modified

What to do when you hit the throughput limits of Azure Storage (Blobs)

In this post we will talk about how we can detect when we hit a throughput limit of Azure Storage and what we can do in that moment. Context If we take a look on Scalability Targets of Azure Storage ( https://azure.microsoft.com/en-us/documentation/articles/storage-scalability-targets/ ) we will observe that the limits are prety high. But, based on our business logic we can end up at this limits. If you create a system that is hitted by a high number of device, you can hit easily the total number of requests rate that can be done on a Storage Account. This limits on Azure is 20.000 IOPS (entities or messages per second) where (and this is very important) the size of the request is 1KB. Normally, if you make a load tests where 20.000 clients will hit different blobs storages from the same Azure Storage Account, this limits can be reached. How we can detect this problem? From client, we can detect that this limits was reached based on the HTTP error code that is returned by HTTP