Artificial Intelligence: The future of computing
(This is a series of blogs which covers various concepts across the board of Artificial Intelligence)
Artificial Intelligence is the ability of a computer to learn by itself. It does this by learning from the data provided to it.
One way to create artificially intelligent models is Deep Neural Networks.
Deep Neural Network
Our brain has a network of neurons inter-connected which work in a specific pattern to bring out a specific output. The output may be throwing a ball or feeling a particular emotion.
In a similar fashion, neural networks are a network on artificial neurons inter-connected to give a certain output on giving it a particular input.
There are layers of neurons particularly divided into 3 types
- Input Layer
- Hidden Layer(s)
- Output Layer
The input data is given at the input layer. This data is processed and the computation occurs in the hidden layers. Once the computation is done, the output is given by the output layer.
How does this work?
Initially, a set of input and well as the corresponding outputs are given to the neural network. This input is fed to the neural network and the output is checked with the given output.
Based on the comparison of the two outputs, changes in the significance of the value of each neuron connection in the hidden layer (known as the weight of the connection) is made automatically to make it more efficient to the desired output. This process of changing the weights based on the output is known as Back Propagation.
Once the model is trained based on the training set, it is tested using the testing set. Once we get satisfactory results, the model is ready to be deployed.
Conclusion
Thus, Neural Networks are a simple way to find solutions to complex AI problems. The simplest neural network can have just 1 hidden layer while as the complexity goes on increasing, the number of hidden layers also increases. As the number of layers increases, the network gets deeper and deeper. Hence, it is called as Deep Neural Network. Deep Neural Networks are the core of Deep Learning applications too.