Deep Learning To AI Solutions

The AI can get accurate when the cognitive powers of a machine are greater

- Kanu Ratan Butani

It’s the cognitive thinking of a machine and its decisive powers for that scenario to decide if the AI of that machine will provide accurate answers or not. This is achieved hugely by the deep learning mechanism.


Before getting on to what deep learning is, let’s see what Machine Learning is. ML is all about teaching a machine about the different scenarios by feeding it with the database and then with the help of data mining and data processing, it will provide with favourable solutions for other scenarios in the future.

For e.g., let’s consider the Lift/Stairs example with some scenarios that are fed in ML.

“Deep learning is an advanced technique of ML. This is done through a neural network just like the brain neural system. But you might be wondering, how is this artificial neural network possible.” —Kanu Ratan Butani, Software Applicatio­ns Manager

1. Person A is 20 yrs old, office is on 2nd floor, and if he is not late, then he will take stairs.

2. Person A = 20 yrs old, office on 10th floor, and even if he is late or not, he will take lift.

3. Person A = 20 yrs old, office on 2nd floor, but getting late, then he will take lift.

4. Person A = 20 yrs, office = 2nd floor, not getting late, but still takes lift because he is having problem in climbing today.

So, just imagine now that these 4 scenarios are there in the database. Now according to the ML, when the situation arises like this, then the machine can decide itself that a person with 30 yrs, and office on 2nd floor, should actually go through stairs.

In most cases, this will be right, but not in all the scenarios. Let’s take this example,

1. Person A = 20 yrs, office on 2nd floor, not getting late, no health issues, can climb stairs, but yet take a life, just because he is not in the mood of climbing the stairs, simple as that.

Since this type of scenario is not captured by ML, the ML mechanism can’t understand that and therefore it will provide a wrong output based on its previous scenarios. This will keep adding in the ML and in the future this type of scenario may provide right answer.

But for all such situations, ML will have to learn and yet miss some option in future. Therefore, this is not the optimal solution. The solution to this issue is not in DB/ or DB management, or feeding in data, but the solution to this lies in Deep Learning.

Deep Learning

Deep learning is an advanced technique of ML. This is done through a neural network just like the brain neural system. But you might be wondering, how is this artificial neural network possible.

This Artificial Neural Network is achieved by many different and complex mathematic­al calculatio­ns, called Convolutio­n Neural Network (CNN) which is the regularise­d versions of multilayer perceptron­s.

Each layer contains different filters, permutatio­ns, combinatio­ns, formulae, and complex algorithms to suite particular function.

When this comes in to picture, the databases get updated, but after passing through all these filters and algorithms that give deeper perspectiv­es to a scenario.

Now consider the above example again. Now the difference is that all the above scenarios pass through this Deep learning CNN process, and then it may give different results. For example, 1. If we consider time factor, so the deep learning algorithm will add some more interestin­g aspects to it and will give a different output, here’s how.

Person A = 20 yrs, office is on 2nd floor, and he can climb, but if the lift is waiting there he may prefer lift, but if the lift takes lot of time to come, then most probably all the people will prefer stairs, even if he is getting late or not. Only exception will be elderly or any one who has issues climbing the stairs.


Deep Learning mechanism is a way to make our machines more intelligen­t, to grow their cognitive powers and human like thinking. The Combinatio­n of CNN and the DL-ML tuning of databases, serves us with more options and handles a different scenario not previously captured, in a more sophistica­ted and intelligen­t manner. This is a preferable concept to take it further for making any solution AI complaint.

 ??  ??
 ??  ??

Newspapers in English

Newspapers from India