Hi everyone,
Requesting help regarding classification coefficient part. I am new to this openvibe, i am sorry if my question are basic and stupid i apologies
Before going to the online experiment with openvibe, I would like to check the result of offline recorded data which was done by using matlab.
Since, i am using (2 class) i use LDA classification method and "classification processor" block. I did not use "classification trainer" block to extract the configuration file.
I manually saw the classification configuration file with (weights and bias-distance) value.
I can extract this values by using Matlab but i did not understand what is "coefficient probability" in the configuration file.
My question :
1) Whether this Matlab way of extracting the (weights and bias ) value will work with the configuration file ?
2) what is the coefficient probability ?
Any guidelines are highly appreciable, Thanking you in advance.
Regards,
GR
classification coefficient
Re: classification coefficient
Hi,
Sorry for the delay I didn't see your question.
This will work.
All configurations are generated to fit a 2-class specialization of the LDA. If you want more details about the calculation please refer to the code or ask me. To get the classification you need to the following calculation:
Weight.transpose() * Features + Bias
and then compare to 0. If the results is negative you recognize the first class, else, you recognize the second.
In order to get probabilistic results, I implemented a new formula, this formula use a constant with is retained under the name of coefficient probability.
Note, this will change is next version.
Guillaume
Sorry for the delay I didn't see your question.
This will work.
All configurations are generated to fit a 2-class specialization of the LDA. If you want more details about the calculation please refer to the code or ask me. To get the classification you need to the following calculation:
Weight.transpose() * Features + Bias
and then compare to 0. If the results is negative you recognize the first class, else, you recognize the second.
In order to get probabilistic results, I implemented a new formula, this formula use a constant with is retained under the name of coefficient probability.
Note, this will change is next version.
Guillaume
Re: classification coefficient
Hi,
Thank you so much, I will try and if i have any problem will contact you.
Thank you
Regards,
GR
Thank you so much, I will try and if i have any problem will contact you.
Thank you
Regards,
GR
Re: classification coefficient
hi, is it possible to give a formula to get coefficient probability ?
Thanking you in advance.
Regards,
Ganga
Thanking you in advance.
Regards,
Ganga
Re: classification coefficient
Hi,
-1/2 * (Mean0 + Mean1).transpose() * CovarianceMatrix.inverse() * (Mean0 - Mean1)
Guillaume
-1/2 * (Mean0 + Mean1).transpose() * CovarianceMatrix.inverse() * (Mean0 - Mean1)
Guillaume
Re: classification coefficient
merci, Guillaume