Deep Learning
If you reached this topic, you should now have a fair understanding about AI and Machine learning.
Deep Learning is a sub category of Machine learning. This technology enables AI to mimic the neural networks of human brain to differentiate between the patterns, sound and chaos of input data. If we learn AI for our exams, we might just learn the definitions; at the next level understand the concepts behind the definitions; at a further level, we may be able to correlate different concepts and solve mathematical equations. Output of this would be an excellent score in the exams! This is called Deep Learning.
Take an example where there are stack of paintings available. Paintings act as an input to the AI. We now want them categorized into landscapes and portraits without us doing the hard work of telling the machine on how to do it.
A deep learning AI would try to decipher the input data and categorize the same. Just like we the neurons in our brain are hidden from plain sight, we have Hidden Layers in deep learning. The hidden layers are located between the input and output of the algorithms and apply weights to the input data. They are responsible for feature extraction and are termed hidden because they are not observable from system inputs or outputs. The presence of more than one hidden layers is called Deep Learning.
In our example of paintings, the first hidden layer can understand and record the data in terms of color pixel matrix. The second layer may determine the edges of the subject in the painting. Third layer may decode facial or object features. Fourth layer would then decode what the painting is about and share it as output.
The more the number of hidden layers in deep learning, the more accurate would the output be. Traditional neural networks contain only 2 to 3 hidden layers whereas Deep Networks may have as many as 150 layers.
