What is Deep Learning - Explained in General Terms

Deep Learning, Machine Learning and Artificial Intelligence are the terms which are being used interchangeably these days. Suddenly every one is talking about them – irrespective of whether they understand the differences or not! Whether one have been actively following data science or not – everyone would have heard these terms. But no one has a clear understanding of what is Deep Learning. Basically it is subset of Machine Learning which in turn is subset of Artificial Intelligence.

Deep Learning is type of Machine Learning which teaches computers to perform actions by learning from examples. Deep Learning is the key technology being used in Driver-less cars that enables them to recognize a stop sign and differentiate between a person and lamppost.It is the key behind the voice controls in various devices like phone, tablets, TVs and hands-free speakers. Deep Learning is gradually prevailing and getting attention lately. Using it, impossible results have been achieved.

In deep learning the computer model learns to perform classification directly from the data like images, test or sound. Learning on their own deep learning keeps on improving the model's accuracy and sometime even exceed the human efficiency.Models are trained on large set of labeled data and neural network architecture that contains many layers.

Try to understand the concept of deep learning with following example:
How would anyone recognize a square? One would see if it has 4 sides, are they perpendicular, is it closed and if all sides are of same length. Which means that the complex problem is to be divided into features that defines it. Therefore, in machine learning user has to define the features to the model while in Deep learning the model identifies the features on its own.

Deep learning also differs from machine learning based on the amount of data each requires. Deep learning requires enormously large amount of data as compared to machine learning. The efficiency of the model improves with bigger volume of data. With less amount of data machine learning proves successful while training a model.

Moreover deep learning can not be performed on low end machines and requires high-end machines as compared to machine learning. Deep learning algorithms work on GPUs and perform large amount of matrix multiplication operations.

The major advantage of deep learning is on Feature Engineering, in which the useful features needs to be created by an extractor by using the domain knowledge. It is the most time consuming step in Machine Learning. But in deep learning the model itself identifies the new features and reduces the need for creating new ones manually.

However if we talk about processing time, deep learning consume a lot of time in training the model as compared to machine learning. While the execution time of deep learning model is seemingly less compared to traditional machine learning.