Microsoft Azure

In the world of Microsoft Azure, it is the “Azure Machine Learning Studio” that consists of all of the tools you have ever dreamed of in order to create and build an Artificial Intelligence (AI) application. It makes use of a GUI-based approach in order to do this, and it can even integrate with other Microsoft tools, most notably that of Power BI.

The Azure Machine Learning Studio Interactive Workspace

As its name implies, this is an interactive workspace of sorts in which you can feed in gargantuan datasets into your AI application, manipulate it, and then complete an exhaustive analysis of it with many ultra-sophisticated statistical functions and formulas, and even get a glimpse of what the outputs will look like from the AI system that you have just built. This entire process is also technically referred to as the “Machine Learning Pipeline.” The main advantage of this is that everything in this process is visually displayed.

It should be noted that the above process can be repeated over and over again as a so-called “Training Experiment” until the results you are seeking have been achieved. Once this has been done, this exercise can then be converted over into the production environment, which is known as the “Predictive Experiment.”

The Machine Learning Studio consists of the following functionalities:

■ Projects:

These are a collection of both the Training Experiment and the Predictive Experiment.

■ Experiments:

These are where specific experiments are actually created, revised, launched, and executed.

■ Web Services:

Your production-based experiments can also be converted to specific Web- based services.

■ Notebooks:

The Machine Learning Studio also supports the Jupyter Networks, which is an exclusive service from the AWS.

■ Datasets:

This is where you upload and store your respective datasets that are to be fed into your AI application.

■ Trained Models:

These are the specific AI models that have you have created and thus have been in trained in the Training Experiment or the Predictive Experiment.

It should be noted at this point that there are certain conditions that must be met first before you can start creating and launching AI models and applications. These are as follows:

■ You must have at least one dataset and one module already established;

■ The datasets that you are planning to feed into your AI models/applications can only be connected to their respective modules;

■ Modules can be quickly and easily connected to other models;

■ There must be at least one connection to the datasets that you are planning to feed into the AI models/applications;

■ You must already have preestablished the needed permutations before you can begin any work.

It should be noted at this point that a module is simply an algorithm that can be used to further analyze your datasets. Some of the ones that are already included in the Machine Learning Studio include the following:

1) The ARFF Conversion Module:

This converts a .NET dataset into an Attribute-Relation File Format (aka “ARFF”).

2) The Compute Elementary Statistics Module:

This computes basic statistics, such as RA2, Adjusted RA2, Mean, Mode, Median, Standard Deviation, etc.

3) Various Multiple Regression Models:

You have a wide range of statistical models that you can already choose from, without creating anything from scratch.

4) The Scoring Model:

This can quantitatively score your Multiple Regression Model that you plan to use for your AI application.

The Azure Machine Learning Service

This is another large platform of Azure which allows your AI applications to be much more scalable. It supports the Python source code, which is the programming language of choice for most typical AI applications. It also makes use of Docker Containers as well. It can be accessed from two different avenues, which are as follows:

■ The Software Development Kit (SDK);

■ Any other type of visual-based interface, primarily that of the Microsoft Visual Studio.

The primary differences between the Azure Machine Learning Services and the Azure Machine Learning Studio are outlined in the following matrix:

Azure Machine Learning Services

Azure Machine Learning Studio

It supports a Hybrid Environment of the Cloud and On Premises

Only standard experiments can be created, launched, and executed

You can make use of different frameworks and instances of Virtual Machines

It is a fully managed by Azure

Azure Machine Learning Services

Azure Machine Learning Studio

It supports Automated Hyperparameter Tuning

It is only available in the Cloud, not as an On Premises solution

The Azure Cognitive Services

This specific service has the following components to it:

1) The Decision Service:

As you deploy your various AI applications, certain recommendations will be provided by this system so that to you can make better decisions as to how to further improve the efficiency and optimization of your AI application.

2) The Vision Service:

This can auto-enable your AI application so that it can analyze and manipulate images and videos.

3) The Search Service:

You can incorporate the Bing Search Engine into your AI application.

4) The Speech Service:

This can convert any spoken words into text format. It also fully supports the Biometric modality of Speech Recognition.

5) The Language Sendee:

This is the Natural Language Processing (NLP) component of Azure, and it can quickly and easily analyze the sentiment of anything that has been communicated, especially those used in chatbots.

 
Source
< Prev   CONTENTS   Source   Next >