REGISTER | LOGIN
Breaking News
News & Analysis

IBM Boosts Brain Understanding

Neurogenesis model beats deep learning in changeable environments
8/30/2017 00:01 AM EDT
5 comments
NO RATINGS
More Related Links
View Comments: Newest First | Oldest First | Threaded View
Les_Slater
User Rank
Author
Re: Sparse neural networks?
Les_Slater   8/31/2017 9:00:15 AM
NO RATINGS
And a network being less sparse, more bulky, due to fuzziness of experence, as repeatedly used more successfully, becomes more efficient in sorting that experience and uses fewer nuerons, the less used atrophying, becoming more sparse.

Les_Slater
User Rank
Author
Re: Sparse neural networks?
Les_Slater   8/31/2017 8:51:34 AM
NO RATINGS
Still not clear to me. I see on some level, sort of a first approximation, an undiferentiated mass of neurons distributed throughout hippocampus, where sparsity is defined as the uniqueness of a network, its simplicity, hence it lack of bulk, number of nuerons and therefore it may be defined by its content or function, a small part of hippocampus bulk.

R_Colin_Johnson
User Rank
Author
Re: Sparse neural networks?
R_Colin_Johnson   8/31/2017 8:38:21 AM
NO RATINGS
Yes, sparse as opposed to a fully populated neural network (which is the usual kind). The neural networks in the hippocampus (and elsewhere in the brain) are physically sparse. The mechanism that makes them that way is neurogenesis, in the hippocampus (at least). Neurogenesis is the process of giving birth to new neurons to process improved encodings and the death by atrophe of the neurons that were set up to process the old obsoleted encoding.

Les_Slater
User Rank
Author
Sparse neural networks?
Les_Slater   8/31/2017 8:32:14 AM
NO RATINGS
Are we talking about physically sparse here, like a particular neural network, as definded by some learned atribute, as widely spaced throughout the hippocampus?

R_Colin_Johnson
User Rank
Author
IBM's Biology Department
R_Colin_Johnson   8/30/2017 4:23:01 PM
NO RATINGS
Who knew that IBM had a biology department? First and foremost it is searching for cures to cancer, and finding them according to Watson. But now its investigating the most subtle functions of the brian to beat the onslaught of deep-learning algorithms with a sophistication that only basic research--IBM's specialty--followed by concrete applications that beat all competitors, but can be licensed for a price. How many times can IBM reinvent itself--from typewriters to big iron to mainframes to PCs to software-as-service to cloud-based everything to artificial intelligence. Did I miss anything? What other science research departments does it have?

Most Recent Comments
Tim R Johnson
 
ewertz
 
antedeluvian
 
ewertz
 
perl_geek
 
R_Colin_Johnson
 
perl_geek
 
R_Colin_Johnson
 
dt_hayden
Like Us on Facebook
EE Times on Twitter
EE Times Twitter Feed