Deep Learning

Top Challenges of the industries faced in handling complexity between data and decision (Making) together

Deep Learning. Artificial Intelligence. Neural networks. Backpropagation. Feed Forward Pass. (Hierarchical-Deep) models. :

Every day we get to hear these fancy terms in one way or another. Although these terms are inter-related but factors which make Deep Learning gain a bleeding edge are quite significant. When the subject at hand has a constantly changing behavior and the deeper understanding of that changing behavior is of utmost importance, Deep Learning could help unlock future insights by using hindsight.  We look at Deep Learning from the lens of innovation, innovation how data is being stored now a days, innovation where techniques gives instant results, Innovation where advanced recommendation systems gives you instant results and it’s something that has definitely seized our curiosity recently. For every organization, Deep Learning is an area of active research these days, especially when you want to be ahead from the competition.

Moreover, Deep Learning methods are beating out traditional machine learning approaches on virtually every single metric.

Hence curiosity is, what exactly is Deep Learning? How does it work? And most importantly, why should we even care?

What is Deep Learning?

Wikipedia (Deep Learning): Deep Learning is a set of algorithm in machine learning that attempt to model high-level abstractions in data by using architectures composed of multiple non-linear transformation.

Before we dive into Deep Learning, let’s take a step back and talk a little bit about the broader field of “machine learning” and what it means when we say that we’re programming machines to learn.

Sometimes we encounter problems for which it’s really hard to write a computer program to solve. For example, let’s say we wanted to program a computer to recognize hand-written digits, addresses, script and a lot other text.

You could imagine trying to devise a set of rules to distinguish each individual digit. Zeros, for example, are basically one closed loop. But what if the person didn’t perfectly close the loop. Or what if the right top of the loop closes below where the left top of the loop starts?

In this case, we have difficulty differentiating zeros from sixes. We could establish some sort of cutoff, but how would you decide the cutoff in the first place? As you can see, it quickly becomes quite complicated to compile a list of heuristics (i.e., rules and guesses) that accurately classifies handwritten digits.

So instead of trying to write a program, we try to develop an algorithm that a computer can use to look at hundreds or thousands of examples (and the correct answers), and then the computer uses that experience to solve the same problem in new situations. Essentially, our goal is to teach the computer to solve by example, very similar to how we might teach a young child to distinguish a cat from a dog.

When and Where Deep Learning is used?

There are many classes of problems that can be solved using Deep Learning Algorithm. Recognizing objects, understanding concepts, comprehending speech. We don’t know what program to write because we still don’t know how it’s done by our own brains. And even if we did have a good idea about how to do it, the program might be absolutely complicated. Few of the problems and sectors using Deep Learning algorithms are used:

Deep Learning_v2-2

How Deep Learning algorithm is to be developed?

Deep Learning Algorithm is a combination of big data and Machine learning algorithms. In simple words,

Multi-layer feed-forward Neural Network (machine learning) + Distributed processing for big data (memory MapReduce paradigm on distributed data) + intelligent algorithm for accuracy (Regularization – L1/L2, model averaging) combine and form Deep Learning.

One of the use cases includes Deep Learning Algorithms to enhance the Fraud Detection system. Fraud detection is a big challenge in modern automated data analytics. Our preferred solution to fraud detection is to deploy an ensemble of various models, then aggregate the majority votes of them for a single binary decision output (fraud/no-fraud).  Even state-of-the-art random-forest or convolutional deep neural networks on their own could still produce decision that underperforms an ensemble of models. So it makes sense to use multiple models (Deep Learning – Neural Networks, Forward Propagation) than a single one. Below is an example of Neural Network – Fully connected directed graph of neurons:

Deep Learning_v2-3

Bluestar has years of experience in the Big Data space and more specific experience in Decision Science, and now working on evolution of Deep Learning in Banking and Telecom domain. While working with some largest Banking and Telecom clients, we realized how much WE can do with their infinite data.

For more information on how Bluestar will help you, please email us at

print page