Introduction To Deep Learning (Neural Network) And Its Mathematical Computations
Deep Learning is majorly used in voice recognition and image recognition in application development. The Algorithm is designed to interpret the customer’s personalization of what the users may wish to do according to their past preferences and usage history.
What is Deep Learning?
We can actually assume this for understanding this better.
Artificial intelligence is the grandfather of deep learning where Machine learning is considered as the father of deep learning. Both Machine Learning(ML) and Deep Learning(DL) are subsets of AI, but deep learning is the next evolution of both Artificial Intelligence and Machine Learning in the recent decade.
Often, the working of Machine Learning is confused with Deep Learning. But NO MORE! Here you can get an overall view and the working.
Difference between Machine Learning and Deep Learning Process
In Machine Learning, algorithms are created according to the data model that is trained by the programmers. Here is the 7 step working process,
- Data Gathering
- Model Creation
Check these Python Tools For Machine Learning You Must Know if you want to learn more about Machine Learning.
Deep Learning learns through the neural network(artificial NN) that acts as a brain and allows the machine to analyze data structure, based on what the brain thinks and does.
How does the Neural Network work?
Neural Network interprets the given data. With the Neural circuit, same does the brain. The artificial neural network will do a better performance when comparing our brain based on the input data. This basic computation of a single neural network is called a Neuron.
Feedforward Neural Network is the simplest artificial neural network and it contains multiple nodes(Neurons) assigned in the layers. Information moves forward in the order of,
Layers of Neural Network
The Input x(i) provides data from the real-time object (world) to the network and in the form of “Input Layer”. There is no computation is processed in any of the input nodes. The input layer passes the data to the hidden layer.
A list of hidden nodes forms a “Hidden Layer”. The Hidden h(i) depends upon the input layer, but there will be no interaction with the real-time object or data ( that’s why is as name “hidden”). It performs mathematical computations and processes the input data to the output layer.
The Output (y) is responsible for computations and takes the information forward from the NN to the real world. It is likely to be called as the multi-layer perceptron.
Neural Networks consist of the following components:
- Weights (w)
- Biases (b)
- Input (x)
- Activation Function (f(x))
- Weight is a connection between neurons that carries a value.
- The bias is also a weight. Every neuron has bias but not in input.
- Calculating the predicted output y, known as feedforward
- Updating the weights and biases, known as backpropagation
Additionally, Major deep learning concepts achieved by,
A Convolutional Neural Network(CNN) is a neural network. CNN is majorly used for image classification and applied to recognize images by the way of convolution.
Types of Layer
- Input Layer
- Convolution layer,
- Activation function layer,
- Pool layer,
- Fully connected layer.
Recurrent Neural Network(RNN) is designed to recognize sequences, for example, a speech signal or a text.
Hopefully, you must have got a clear idea of deep learning achieved through Neural Network and how it works from my above explanation. Now, try learning how to build a simple neural network in 9 lines of Python code for practical implementation.
Understanding the basics of this process can be helpful when working with Neural Network (NN). I recommend you to try to program a small neural network, only with Numpy.