If you’re living under a rock, Deep learning is a sub-branch or an approach to machine learning. DL is inspired by mostly what we know about the brain. So it's very obvious, its only and not so closest (maybe) goal is to create a neural network model just like our very own.
In ML, Linear Regression, Logistics Regression, etc. are the traditional algorithm based on Statistical Analysis are used. Processing Big Data is one of the major limitations of the Sklearn library because it works traditionally on Statistical algorithms. To automate the feature selection and to process the bigdata, Neural Networks are used. Neural Networks is also known as Artificial Neural Networks which is inspired by the biological neural networks that constitute human brains.
The network starts with the input layer and ends with the output layer (Layer of Choice), in between are the hidden layers, the number of layers in this hidden layer depends upon the model and use case. Neural Network is based on a collection of connected units or nodes/neurons. Each neuron store a number/ data, which are activations. Activations of one layer are activations of the next layer and so on until the output layer. The real core mechanism of the NN is exactly how the activations bring about the activation in the other layer.
In the biological network of the neuron, some group of neurons firing signals causes certain other neurons to fire and so on. This process is training, done repeatedly forms a pattern. The more the data, the better-trained model we have. Here accuracy of the model corresponds to the intelligence of the biological brain.
And According to Leo Tolstoy,
“Everything intelligent is so boring”
Well that’s true because training a model takes a huge amount of time as compared to the traditional ML model and obviously a whole other level of computing power. Leo definitely was long-sited.
“Most of the research in Deep Learning world is a total waste of time, it’s a problem of Science in general which needs to be published, which means they need to work on things that their peers are extremely familiar and can recognize in advance in that area, so they all need to work on the same thing. So they don’t work on things that are practically useful to get just a whole lot of research and minor advances in the highly studied have no significant practical impact…” -Jeremy Howard (Google Him)
Well, by now you might be wondering that it has so many use cases that probably every industry is harnessing it, you’re partially correct, every institute or company wants to benefit from it. Only 1% of the total industry currently has hands-on DL integrated. The number is pacing up thanks to the hardware improvements. The top applications of DL are
So is it the number of layers, huge amount of data, or its analogy to human psychology (Brain) that it's called Deep Learning? Well somewhat, it’s the huge network of neurons and humongous amount of maths involved, and ever-growing computation that’s called Deep Learning.
If you guys want to learn more about Deep Learning/ AI, you could binge on some courses and Lex Fridman’s AI podcast.
Stay tuned for more brief words.
-Arther
