Deep Learning Explained

by yattish ramhorry

Have you ever wondered how Google translates an entire webpage into another language in a matter of seconds? Or, how your phone gallery groups photographs based on their location? All of this is a product of Deep Learning.

But What Exactly Is Deep Learning?

Deep Learning is a subset of Machine Learning, which in turn is a subset of Artificial Intelligence.

Artificial Intelligence is a technique that enables machines to mimic human behavior, and machine learning is a technique that achieves AI through algorithms trained with data.

Finally, Deep Learning is a type of machine learning that is inspired by the structure of the human brain. In terms of Deep Learning, this structure is called an Artificial Neural Network.

Let’s understand Deep Learning better, and how it is different from Machine Learning.

Assuming that we have a machine that can differentiate between tomatoes and cherries. If this is done using machine learning, we would have to tell the machine the features based on which the two can be differentiated. These features could be the size and the type of stem on the tomato or the cherry.

With Deep Learning, on the other hand, the features are picked out by the neural network, without any human intervention. Of course, that kind of independence comes at the cost of having a much higher volume of data to train our algorithm.

Let’s dive into the working of neural networks!

Supposing that we have three students, and each of them writes down the digit ‘9’ on a piece of paper. Notably, they all don’t write the digit identically. The human brain can easily recognize the digits, but what if a computer were to recognize them?

That’s where Deep Learning comes in. A neural network that is trained to recognize handwritten digits, will present each number as an image of 28 x 28 pixels.

That amounts to 784 pixels. Neurons which are the core entity of a neural network is where the information processing takes place. Each of the 784 pixels is fed to a neuron in the first layer of our neural network. This forms the input layer.

On the opposite end, we have the output layer with each neuron representing a digit with the hidden layers existing between them. The information is transferred from one layer to another over connecting channels. Each of these connecting channels has a value attached to it and hence is called a Weighted Channel.

All neurons have a unique number associated with them called “bias.” This bias is added to the weighted sum of the inputs reaching the neuron, which is then applied to a function known as the “Activation function.”

The result of the activation function determines if a neuron gets activated. Every activated neuron passes on information to the following layers. This continues up till the second last layer. The one neuron that is activated in the output layer corresponds to the input digit.

The weights and biases are continuously adjusted to produce a well-trained network.

Where Is Deep Learning Applied?

Deep learning is applied in customer support. When most people converse with customer support agents, the conversation seems so real, they don’t even realize it’s a Bot on the other end.

Deep learning is also applied in medical care when neural networks are used to detect cancer cells and analyze MRI images to give detailed results.

Self-driving cars are another area that Deep learning is applied. What seems like science fiction is now a reality. Tesla, Apple, and Nissan are just a few of the companies that are working on self-driving cars. So, Deep Learning has a vast scope, but it also faces a few limitations.

Limitations Of Deep Learning

The first limitation of Deep Learning is data. While Deep Learning is the most efficient way of dealing with unstructured data, a neural network requires a massive volume of data to train.

Letts assume that we always have access to the necessary volume of data. Processing this volume of data is not within the capability of every machine.

Which is the second limitation of Deep Learning, Computational power. Training a neural network requires graphical processing units or GPUs which have thousands of cores as compared to a CPU. GPUs are of course more expensive than CPUs.

Finally, we come down to training time. Deep neural networks take hours or even days to train. The time increases with the amount of data and the number of layers in the network.

As these limitations in Deep Neural networks become increasingly redundant, we will find more companies adopting Deep Learning and Deep neural networks to train their data. This leads to better accuracy in predictions and makes deep learning models less error-prone.

With higher accuracy of models and larger volumes of data, the possibilities in Deep Learning applications become endless!