Nt1310 Unit 3 Lab Report

512 Words3 Pages

determine each pixel belongs to background or foreground.Wis the weights between the pattern and summationneurons, which are used to point out with which a pattern belongs to the background or foreground. They areupdated when each new value of a pixel at a certain position received by implementing the following function:Wt+1ib=fc(1−βNpn)Wib+MAtβ!(37)Wt+1i f=(1−Wt+1ib)(38)whereWtibis the weight between theith pattern neuron and the background summation neuron at timet,βisthe learning rate,Npnis the number of the pattern neurons of BNN,fcis the following function:fc(x)1,x>1x,x≤1(39)MAtindicates the neuron with the maximum response (activation potential) at frame t, according to:MAt1,f or neuron with maximum response0,otherwise(40)Function …show more content…

The total frame is :T=Tno+To(41)The weights for the pattern neuron corresponding to the feature value are determined:WTib=(1−βNpn)TnoWoib+Toβ!(42)WTi f=1−WTib(43)whereWoibcorresponds to the initial weight set when the pattern is first observed. …show more content…

If it is not encountered forTnoframes, then the confidence that afeature value belongs to the background will decay from the maximum to(1−β/Npn)Tno.Activation and Replacement Subnets: The activation subnet contains two functions: first it can point outwhich net work has the maximun output and whether the maximum value exceeds the threshold. If the maximumvalue does not exceed the threshold then the background network will be inactivated and the weight of thepattern neuron is considered to be replaced. On the contrary, if exceeded, the feature is considered belonging to aforeground object.In the first layer, a single neron is used to indicate whether the network is activated or not. This layer contains, Vol. 1, No. 1, Article . Publication date: January 2018.

More about Nt1310 Unit 3 Lab Report

Open Document