Login
Password
Sources on this Page

> Headlines by Category

 Home / Science / Mathematics

You are using the plain HTML view, switch to advanced view for a more complete experience.

compute area from contour lines
hi,
contour plots i prepared for 5X15 matrix.....
so, at moment i would like to compute area and volume between each contour line..

can any one help.....

Thank you in advance..
Re: SIMULINK-Doing a element-based FFT
I have considered latency, but the outputs are completely different. My code and outputs are here. Both In1 and In2 are connected to the same signal from the workspace. The MATLab function contains the function bitrevorder(x), which switches back the reversed bits from the FFT. The output ports 2 and 3 lead to a vector scope based on frequency. At an interesting point in time, output port 2 displays the correct frequency peak of about 30, and output port 3 displays a peak at about 230 kHz.

The image to the code is here: https://ibb.co/nO1JQ5
The image to the output port 2 is here: https://ibb.co/bAezXk
The image to the output port 3 is here: https://ibb.co/iUha55

Thank you.
Curve Fitting toolbox plot() error
When I try to use the plot function within the curve fitting tool box I always receive this error:

Undefined variable "curvefitlib" or class "curvefitlib.internal.getLegendItems".

Error in cfit/plot>updatelegend (line 360)
[oldh, str] = curvefitlib.internal.getLegendItems( gca );

Error in cfit/plot (line 251)
   updatelegend(legh, legtxt);


Does not matter what code or data I use, even comes up when I tried MATLAB's plot() example:
https://au.mathworks.com/help/curvefit/plot.html

I tried, ls(fullfile(toolboxdir('shared'),'curvefitlib'))
+curvefitlib shows up


Any Ideas what the problem could be?


Using R2016b mac
Re: Specifying number neural units of each hidden layers in Neural Network
Hien Quang Bui" wrote in message <oj15gc$29o$1@newscl01ah.mathworks.com>...
> Thank you!
> https://www.mathworks.com/help/nnet/ug/design-time-series-narx-feedback-neural-networks.html
> Beside, I can not match the result of NAXR model after finished train data. One of reason is that the resolution of figure is low quality image. Could you please provide the figure with high resolution?
> Hidden layer is 2 and delay is 2. How can we rewrite the equation?

1. Do not write above previous posts; write below them.
2. Are you talking about the documentation article referenced above:

Design Time Series NARX Feedback Neural Networks ?

If so, there is no image... just plots. The data in the reference is in the
magdata file. Just down load it and plot it yourself.

3. If there were images, why do you think I could provide a clearer copy
than you could obtain yourself?

4. In 27 years of designing NNs, I don't recall using more than 1 hidden layer
except when classes are represented by multiple (Gaussian) clusters.

Greg
Re: SIMULINK-Doing a element-based FFT
Navan Ruthramoorthy" wrote in message <oj0hd9$hpo$1@newscl01ah.mathworks.com>...
> "Daniel Yang" wrote in message <oivfhh$42j$1@newscl01ah.mathworks.com>...
> > In order to convert simulink into HDL code, I cannot use the buffer block, and thus, must do a element-based fft. Unfortunately, no matter how I try, I cannot get the sample-based fft to work (using the HDL-optimized FFT), and the FFT only works when I do it on a frame. My first question is as follows.
> >
> > Theoretically, doing a sample-based FFT, then creating a buffer should be the same as buffering, then doing the FFT on the frame. However, these return different outputs. Why is this?
> >
> > Secondly, the Spectrum Analyzer achieves the desired result flawlessly (whether using frame-based or sample-based input). Is there a way to find the spectrum analyzer code, or is it possible to reveal how the Spectrum Analyzer is coded?
> >
> > Thanks in advance!
>
> Any specific reason why you cannot use a buffer? HDL-optimized FFT will have latency in its output. When you compare your results you need to account for this latency. Except for latency it should match the FFT results. I would expect Spectrum Analyzer to have an internal buffer and then compute the spectrum when the buffer is full.
>
> Navan



Hello,
Thank you for your fast response. I may be naive, but buffers don't work for HDL generation. While checking compatibility, the software states that the block is not compatible for HDL generation. Is there a way to fix this, or another block that does what a buffer does.

Thanks.
Super-strong metal made for next tech frontier
Engineers have developed a strong, durable new material to help shape advanced MEMS sensors needed for the internet of things.
Re: Specifying number neural units of each hidden layers in Neural Network
Thank you!
https://www.mathworks.com/help/nnet/ug/design-time-series-narx-feedback-neural-networks.html
Beside, I can not match the result of NAXR model after finished train data. One of reason is that the resolution of figure is low quality image. Could you please provide the figure with high resolution?
Hidden layer is 2 and delay is 2. How can we rewrite the equation?

"Greg Heath" wrote in message <oinls9$4bs$1@newscl01ah.mathworks.com>...
> "Hien Quang Bui" wrote in message <oij0qr$hrf$1@newscl01ah.mathworks.com>...
> > Dear Mr. Greg,
> >
> > I use app to generate NARX model. How can I specify number neural units of each hidden layers in Neural Network?
> >
> > Input = 4
> >
> > Hidden layer1: 3 neural units
> >
> > Hidden layer2: 3 neural units
> >
> > Output=1
> >
> > Thank you very much!
> > Best,
> > Hien Bui
>
> 1. I never use more than 1 hidden layer
> 2. I use divideblock to ensure only predicting the future with the past
> 3. I minimize H subject to the constraint that mse(etrn) <= 0.01 var(ttrn,1)
>
> Details are in my NEWSGROUP and ANSWERS posts. Search on
>
> greg narxnet
>
> Hope this helps.
>
> Greg
Re: How to analyse the results of training of neural network ?
1. Calc MSE00 after I N O
2. 9 input outlier search with mapstd
Tabulate (1:9) min med mean std max
3. Plot data. Include above stat lines if helpful
4. Modify or delete outliers
5. Assign trueclasses before loops
6. Tabulate NMSE and PCTERR inside loops
7 Try dH = Hub/10;
8. If above is insufficient, try H > Hub. Validation stopping will prevent overtraining the overfit nets.
9. Tabulate results in matrices
10. For NEWSGROUP/ANSWERS references search

                  PATTERNNET NTRIALS

Good Luck

Greg
Engineers design a robotic gripper for cleaning up space debris
Researchers combined gecko-inspired adhesives and a custom robotic gripper to create a device for grabbing space debris. They tested their gripper in multiple zero gravity settings, including the International Space Station.
Re: printing unwanted black background
Henry" wrote in message <ln6vh2$fph$1@newscl01ah.mathworks.com>...
> "Sasan " <markjohnson@yahoo.com> wrote in message <j8fsdb$qo5$1@newscl01ah.mathworks.com>...
> > Whenever I try to print a figure the output (jpeg, png) always has a black background, even when I set a white or clear background. My figures utilize the pcolor, surf, and/or slice functions. If anyone has encountered such issues and has found a solution, please share.
>
>
> So I have had this problem when every figure i exported using the openGL renderer has a black background which i did not want. I searched through the internet for a solution but never came across an answer. I thought I would post on here in case anyone else comes across this thread.
>
> It appears that a setting has been unwillingly changed (although I dont know which one). My solution was to find the preferences directory by typing prefdir into the command line.
>
> prefdir
>
> Close matlab
>
> Delete this folder
>
> Re-open matlab.
>
> The folder is recreated but not previous settings or history is used by matlab. It is quite annoying but it solved the issue with always having the black background with exported figures

I narrowed the issue down to the matlab.prf file in the folder you mentioned. If you close MatLab, just delete or rename that one file, then restart MatLab, it creates a new file and fixes the issue while limiting the collateral damage (losing the command and working directory history and probably other things). The file is human readable with some of the lines being search and replace history from the text editor, but I could not figure out what other lines meant. I also compared the full property details of the figure with and without the problem and could not find any difference.

Only possible things I could think of that would change these preferences are that I had recently changed the default printer from OneNote to a real printer, and I had recently undocked my laptop and connected to a different external monitor which seems to give MatLab graphics drivers grief sometimes.

Thanks for the original solution post.
Post Selected Items to:

Showing 10 items of 572

home  •   advertising  •   terms of service  •   privacy  •   about us  •   contact us  •   press release design by Popshop •   © 1999-2017 NewsKnowledge