Welcome to “What is Deep Learning?” After watching this video, you will be able to: Define key concepts in deep learning. Explain how deep learning fits into the machine learning ecosystem. Describe how deep learning works. List the layers of the neurons. And, list some common deep learning applications. Deep learning is a subset of machine learning that can be defined as an approach to machine learning using past data based on mathematical models that mimic the human brain. Deep learning eliminates some of the need for pre-processing data and automates feature extraction. For example, let’s say you have photos of foods sold at a truck stand like hot dogs, burgers, fries, and so on. Deep learning can determine which features are most important for identifying each one. This determination is done by the computer rather than a human. Deep learning can be applied to different kinds of data, whether tabular (that is, data in rows and columns) or non-tabular (that is, images, videos, and text). Deep learning requires large amounts of data. In deep learning, you will come across concepts that are used every day in the field. The first one is artificial neural network, or ANN. An artificial neural network is a group of connected nodes called neurons laid out in one or more layers, loosely modeling the human brain. A neuron is a computation unit that can be expressed as a weighted sum of inputs. A simple case is shown here, It takes in an input X 1 and assigns a corresponding weight W 1, then an input X 2 and a corresponding weight W 2 and an input X 3 and a corresponding weight W 3. This can go all the way to X N and a corresponding weight W N. The configuration of all the inputs is called a perceptron. The perceptron also contains a summation function which is a total summation of the inputs and their associated weights. Each neuron is associated with an activation function that predicts the value of the output. A neural network with one neuron is called a single-layer neural network. When a neural network has an input layer; one processing layer, commonly called hidden layer; and an output layer; it is called a multilayer perceptron. When an artificial neural network has two or more hidden layers, it is called a deep neural network. The neurons perform their computations on the input data and predict an output based on the activation function. An activation function is a mathematical equation that determines the output of a neural network. The function is attached to each neuron in the network, and determines whether the neuron should be activated or not, based on whether each neuron’s input is relevant for the model’s prediction and provides an output for the next layer. Activation functions also help normalize the output of each neuron to a range between 1 and 0 or between -1 and 1. Deep learning mimics the human brain through an artificial neural network, which is a mathematical model that uses a combination of data inputs, weights, and bias. Like the brain, the model also deals with neurons. Let’s assume you wanted to predict the price of a house. The neural network works in three types of interconnected layers starting with the input layer, which receives the independent variables, for example, number of rooms, location, and age. The input layer passes the input to the first hidden layer. Other hidden layers process the inputs by assigning weights that determine the importance of each input value. And then the output value gives the prediction. Deep learning is used in a lot of applications. We have Self-driving cars – Deep learning is the backbone of autonomous cars. The deep learning algorithm allows the cars to keep learning by collecting data through their sensors and cameras and interpreting it. Deep learning is also commonly used in developing classifiers that detect fake news and filter it out of your feed. Another area of application is natural language processing. Deep learning can catch linguistic nuances and frame the appropriate responses to the user. Examples include virtual assistants like Alexa and Siri. Deep learning is also used to build computers that can play games. For example, this is a simple case of a computer using deep learning to play “rock, paper, scissors” with a person. In this video, you learned that: Deep learning is loosely based on the human brain. Deep learning is a subset of machine learning. There are three layers in the neurons: the input, hidden, and output layers. Neurons use an activation function to normalize the outputs. And, deep learning is used in a lot of applications, for example, self-driving cars, fake news detection, and natural language processing.