|
|
|||||
|
|
|||||
|
The matrices for generating electrical
stimuli and collecting responses from neuronal cultures are increasingly
evolved and miniaturized. The most ambitious goals foresee the
use of neuronal cultures for data storage. We are still far from these goals
and there are innumerable problems that need to be solved in order to obtain
this result in a practically usable way. One of the fundamental steps is the
ability to distinguish two different neuronal cultures on the basis of the
electrical stimulus and the electrical response they give on special electrode
arrays. One of the possible approaches to obtain such discrimination can be
the analysis of stimulus and response images on the matrices using Deep
Learning algorithms. The disadvantage of this approach is that each new
dataset requires a retake of all previously acquired data. We have used Mythos™ algorithm which
analyzes tens of thousands of stimulus-response matrices but has the ability
to learn new stimulus-response associations in real time without cycling on
previously acquired data. The Proof Of Concept was based on the simulation of
the response of neuronal cultures through pseudo-random sequences whose seed
was associated to each particular neural culture. |
|
|
|
©2024_Luca_Marchese_All_Rights_Reserved Aerospace_&_Defence_Machine_Learning_Company VAT:_IT0267070992 NATO_CAGE_CODE:_AK845 Email:_luca.marchese@synaptics.org |
|
Contacts_and_Social_Media
|
