Deep Learning

Human vs. machine

Humans and machines (=computers) have different capabilities. Dings being difficult for us like calcalations, learning poems or process algorthims are very easy for machines. In constrast it's very easy for humans to detect objects like persons, animals or text on images, driving cars or forecast customer wishes. These things are very difficult for computers.

  Human Machine
Calculations - +
Learn poems - +
Process algorithms - +
Reading text + -
Detect objects on images + -
Driving cars + -
Forecast customer wishes + -


Learning vs. programming

Learning and programming are two concepts for solving problems. Many things that are 'easy' for humans needed lots of cumbersome learning firstly. Remember how we learned reading in school. Before school we may have some characters already, but only uppercase ones in block letters. While leaking the characters, we did not know the horizontal lines of the paper.

We've solved the 'E' problem by training it again and again. Some years later we were able to read other fonts.

During programming the propramme creates a program that the computer processes. This program is very fixed and normally processes inputs to outputs. No experience is needed for this and it can be very fast. Problems like calculating or storing text can easily be solved.

Neural Networks

Human learnings happening with the help of neural networks in brain in which neurons are connected with nerves. During learning these connections are established. Artifial neural net try to clone this concept in the memory of a computer. A neural net is programmed, the connection are based on training the net using data. This training is called Deep Learning.