So, what is deep learning?

[sumaiyakhatun27]

In fact, solving such a problem is an extremely difficult thing, because it is difficult to formalize. You can create an algorithm that will recognize cats in certain types of photographs (with a similar angle, with similar sizes of objects). You can create an algorithm that will draw these cats. But how can you create an algorithm that will recognize cats from any angle if it, the algorithm, does not understand what a cat is? The neural network had to start understanding this.


And it learned it. It learned it itself, a person did not set it such a task. netherlands email list It is clear that everything is not so rosy yet. The neural network distinguishes cats in images only in 15% of cases. The accuracy can be much higher if the network studies similar materials. That is, in fact, the accuracy of cat recognition by the neural network is currently much worse than the recognition of cats by a 4-5 year old child.


But the neural network, unlike the child, was not trained by anyone, no one showed objects with a cat, naming them. The network gave birth to the concept of a cat on its own. WE WILL PROMOTE YOUR BUSINESS Read more What will happen next? In March 2015, a "nuclear bomb" exploded in the SEO world - Google published an article " Trust Based on Knowledge: Assessing the Reliability of Web Resources ".