To do this, they changed the time constant for cells in the neural network - in other words, how quickly the cell reacts to other cells connected to it. First, the researchers associated the response time of each cell to a single value. They then compared this case to what would happen if the response time was a randomized value chosen from a gamma distribution. In this case, some cells reacted immediately, while others took some time and “waited” to see what nearby cells did.
In fact, these randomized neural nets were able to perform just as well, or better, than those with more cells. “Heterogeneous networks have no additional cost in terms of the number of neurons or synapses,” say the authors.This heterogeneity is similar to that seen in human and animal brains.
“We cannot rule this out,” say the authors. Yet, it is possible that brains evolved to be like this. “Our results show that the distributions observed experimentally closely match the optimal ones found by simulations, and this optimal distribution confers a substantial computational advantage.”getty
Therefore, it could be very likely that brains evolved to be efficient by being different. Brains with heterogeneous neurons are more robust, and their differences help them to learn in a variety of environments. These differences may be part of the reason that humans and animals learn so well, and so quickly.
These results can help build a better-performing AI system. By allowing each cell to vary slightly, neural nets can be smaller and easier to build.
There is no greater joy than having financial stability and a life free of debts, that’s why I keep tweeting about lisamiaBTC it has been a pleasure collaborating with you...there is no other way I can show my appreciation other than tweeting the good deed you have done
... The error is trying to view AI systems as whole machines with a specific purpose. Life has no purpose but to exist. To replicate Life, or create it, (ethically), we would have to operate under the assumption every constituent could organically fit in the environment.
thanks
imperfectly perfect
United States Latest News, United States Headlines
Similar News:You can also read news stories similar to this one that we have collected from other news sources.
Source: Forbes - 🏆 394. / 53 Read more »
Source: Forbes - 🏆 394. / 53 Read more »
Source: Newsweek - 🏆 468. / 52 Read more »
Source: Reuters - 🏆 2. / 97 Read more »
Source: Forbes - 🏆 394. / 53 Read more »