Hola chicos,

this weekend I will be a participant of a hackathon!!! Yey, right? Honestly, I am super excited but I feel a bit unprepared. Hackathon challenge is going to be about image processing so there is a lot to learn about it. I’m not sure whether the problem will be given as supervised or unsupervised but what I do know – from my brief research, that I need to be acquainted with deep learning so today I want to focus on that.

Question for today friends: what deep learning is about?

The concept of deep learning has over 50 years as for now, but until the second decade of the 21^st century, there was not enough data to use it and not strong enough processors to perform it. The main idea behind it is to process or analyze the problem as the human brain does, which is process it simultaneously within numerous neurons connected with each other. And from that objective programming systems and structures (often machine learning processes) operating on that idea are called neural networks. So deep learning is a part of machine learning which is a part of artificial intelligence.

Following question arises: what is a neural network?

This is a enormous number of connected with each other and working at the same time processors. Each processor has it’s own memory and task to do, which often might be a machine learning process. The network learns something simple and basic at the beginning of the hierarchy and then passes information about it’s discovery to the next level. There, basic information is combined into something more complicated and so on, until information is processed through all hierarchy and through all processors.

Difference between Machine Learning and Deep Learning:

As previously stated, deep learning is a part of machine learning. But machine learning always needs to have a human operator for providing testing examples and manually amend mistakes. What is more machine learning is very time-consuming and it only depends on human creativity when figuring out how to make a model better.

On the contrary, deep learning is mostly perform as unsupervised learning. Raw data are provided for the model and no more human participation is needed, as computer thinks independently and learns. In deep learning, we are not using linear logic, as in most machine learning problems, but layers allow the model to learn and they reorganize after every new experience.

This is all for now, thanks!!

xoxo,

szarki9