Home > Nia > Blogs > Why Infosys Nia™ Advanced Machine Learning (part 1)

Why Infosys Nia™ Advanced Machine Learning (Part 1)

July 13, 2017 - Nicholas Martin Ball

Machine learning is often mentioned in the same breath as Artificial Intelligence (AI) these days. There are many who lay claim to the uniqueness of their AI offering because it features machine learning (ML). So what sets Infosys Nia apart? Why is Infosys Nia Advanced Machine Learning better positioned to deliver success for businesses? Let us explore.

In March 2017 Infosys acquired Skytree, The Machine Learning Company. The capabilities of Skytree are now found integrated within the Infosys AI platform, Infosys Nia, and are now known as Infosys Nia Advanced Machine Learning. There are three major aspects that make it uniquely effective:

  • Automation
  • Speed & Scalability
  • Ease of Use

Let us examine these briefly.


Typically in data science, to create a good ML model, the user has to choose which ML algorithm to run given their data (decision tree, deep learning, support vector machine, etc.), then select suitable values and optimally tune any of up to dozens of hyperparameters specific to that algorithm. The only way to do this well is to be an expert.

In contrast, the Infosys Nia AutoModel capability allows a user who doesn’t know any of this to still build ML models, or allows the expert to save a great deal of time that would have been spent in tedious manual model tuning. AutoModel combines our Smart Search through the algorithm hyperparameter space with the ability to tune multiple algorithms in one run, resulting in a model that is competitive with the experts but with very little user effort or time. The AutoModel process is fully general, updating both the algorithm chosen and the hyperparameters tested after every iteration.

In addition to AutoModel, we have further generalized the system to automatically perform feature engineering on an input dataset; this is our functionality called AutoFeaturize. As is well-known, good feature engineering can significantly improve the value of a model, but it can be time consuming. With Infosys Nia, further user time is saved by not having to perform all such engineering manually. Feature engineering is a huge field, and we are working to rapidly expand this capability in future releases.

Speed & Scalability

Infosys Nia Advanced ML is written from the ground-up for speed and scalability. We have benchmarked our algorithms against other software, both commercial and state-of-the-art open source, and have never lost in speed, or in model accuracy, including in all customer proof-of-concept (POC) engagements.

This focus on speed allows users to extract maximum accuracy while using all of their data, a capability that has been worth many millions of dollars to some of our customers.

As a quantitative example, in our command line system we have run gradient boosted decision trees on a 1 trillion element training set (10 billion rows × 100 columns) on a regular Hadoop system, demonstrating near-ideal weak and strong scaling from a single node up to 100 nodes. In a second example, we ran a training set of approximately 500 million rows × 50 columns in the graphical user interface (GUI), giving scale combined with all of our ease-of-use features.

Ease of Use

Infosys Nia Advanced ML is designed to give you the full power and business value of advanced machine learning while remaining easy to use. Through the GUI you can access the full automation and speed & scale capabilities of the system, without having to write any code. For those users who prefer to write their own code, a programmatic interface is available as an SDK using a binding language such as Python. Running your code, for example in a Jupyter notebook, causes the project to simultaneously appear in the GUI, thus allowing real-time interaction. For those who want closer integration with other tools, the underlying API is also available.

The platform effect

Additionally, Infosys Nia Advanced ML can leverage the capabilities of other parts of the Infosys Nia platform. For example, Infosys Nia Data’s extensive set of connectors to data sources such as HDFS, databases, social media, and so on. The platform is also extensible to include outside open source tools, thus also extending the reach of Infosys Nia Advanced ML.

Now it might be said that there are other offerings in the market which may contain aspects of the capabilities of Infosys Nia Advanced ML discussed here. But what is highly unlikely to be found is the combination of all three of these qualities: automation, speed & scalability, and ease of use, in a single product. Additionally, the advanced features unique to Infosys Nia result in a uniquely capable offering designed to appeal to both expert data scientist users and the much wider community of business users and analysts. Therein lies the significance of these coming together within an integrated platform.

Related Blogs All Blogs


What did the buffalo say to his son when...
April 27, 2017


Do AI technologies question the basis of humanity?
May 17, 2018

Leave a Reply

Your email address will not be published. Required fields are marked *