innovator

Add idea


Calendar

«    August 2019    »
MonTueWedThuFriSatSun
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
 

 

Advert

 

Payment

 

Advert

 

Authorization

Стартап
DateDate: 13-08-2019, 07:03

Using the results of advanced brain research, a group of scientists from the University. Bar-Ilana in Israel has revealed a new type of ultrafast artificial intelligence learning algorithms based on slow brain dynamics. They partially exceed performance indicators achieved with the most advanced AI learning algorithms.
“The modern main scientific and technical point of view is that neurobiology and machine learning are two different disciplines that develop independently,” said the lead author of the study, Professor Ido Kanter. “But we are perplexed why these two areas do not develop in parallel.”
They added that the number of neurons in the brain is less than the number of bits in a typical drive of modern personal computers, and the computational speed of the brain is much less. However, the rules for brain training are very complex and far from the principles of teaching modern AI algorithms.
Brain dynamics do not comply with well-defined rules synchronized for all nerve cells. For example, the brain and human vision can read the location of several objects at once. At the same time, traditional artificial intelligence algorithms are based on synchronous inputs, so the AI cannot do the same.
A new study shows that ultra-fast learning speeds are equally effective for small and large networks. Therefore, according to the researchers, “the lack of a complex brain training scheme is actually an advantage.” Another important discovery is that learning can occur without learning steps through self-adaptation in accordance with asynchronous inputs.
Source: hightech.fm





Add comment

Your name:*
Your E-Mail:*
Input code
from the picture*
Enable this picture to display the security code
reload, if the code cannot be seen