Patternnet for 14 -class classification problem

2 次查看（过去 30 天）
Gmoj Goo
Gmoj Goo 2015-3-24
评论： Greg Heath 2015-5-6
Hi,
I am trying to classify the number of 0 and 1 appearing after a binary time series. What i have is a very large sequence of [ 0 1 1 0 1 ] which precedes say a continuous sequence of 3 zeros. ie [ 0 1 1 0 1 0 0 0 ]. I want to use the patternet to do this ie input [0 1 1 0 1] and the network should classify it as 3. From the binary series i have extracted the Input and the corresponding target. I want to classify upto 7 continuous zeros (1 class for 1 zero, 2nd class for 2 continuous 0s and so) and ones (8th class for 1 ones,..14th class for 7+ ones). ie 14 classifications. The problem is that it is not able to do so accurately. There is a confusion of more that 0.8. It classifies most as 5, 6 (zeros of length 5 and 6) and 14 - ones of length 7 or more (the most). Other classes aren't classified.
I am using the default parameters (trainscg, max_fails =6) and for hidden layers i have [20 30]. I generate around 30k Input Target values. Train:Validate:test = 0.6:0.2:0.2. I am using eye(14) to create the targets.
My question is - Is what i am trying to do even possible using patternet? If yes, what are the things i need to take care of so i get better results. and if no which network should i use?
Thank you very much.
  3 个评论
显示 1更早的评论
Gmoj Goo
Gmoj Goo 2015-3-25
编辑：Gmoj Goo 2015-3-26
size(input) = [I N ] = 30*28008
size(target) = [ O N] = 14*28008
inputexample = [2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 1 1 1 1 1 1; .. ]'
targetexample = [0 0 0 0 0 0 0 0 0 0 0 0 0 1; ..]'
I am using eye(14) to create the target values. Here I have replaced the [0 1] with [1 2] in the input.
Greg Heath
Greg Heath 2015-5-6
I do not understand why that target is associated with that input.

请先登录，再进行评论。

请先登录，再回答此问题。

回答（0 个）

类别

在 Help Center 和 File Exchange 中查找有关 Pattern Recognition and Classification 的更多信息

标签

产品

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by