Norman Sasono

Beauty is the first test: there is no permanent place in the world for ugly Mathematics – G.H. Hardy
See also: Other Geeks@INDC

About Me

Probabilistic Neural Networks works better for classification problems

If you have checked out my paper about the using of Feed-Forward Backpropagation Neural Networks, here's another approach that I used. It is called the Probabilistic Neural Networks. This approach can forecast the onset of Diabetes Mellitus with 100% accuracy. Compared to the Feed-Forward Backpropagation approach that could give no more than 93%.

Probabilistic Neural Networks was actually designed for such classification problems. No wonder.

Share this post: | | | |
Posted: May 02 2008, 04:30 PM by norman | with 4 comment(s) |
Filed under: , ,

Comments

arifin_tjia said:

Did you know? that what kind of the algorithm in text mining and what they are?

I felt confuse in looking for algorithm in text mining thank before.

# May 6, 2008 11:02 AM

adhiguna said:

Great mas Norman, saya akan coba algoritma ini untuk aplikasi machine vision. Thanks untuk insightnya...

# May 14, 2008 3:13 AM

acoltye said:

back to matlab again boss ?!

sayangnya matlab 12 semakin sucks gara2 make Java,

berat dan boros memory (mempry limit up to 1.7 GB for 32 bit machine) ...

trus boss... hati2 dengan classifier yang kasih output akurasi 100%! itu kadang2 bukan indikasi yang baik, sebab bisa jadi classifier NN itu mengalami overfitting akibat overtraining pada dataset.

en.wikipedia.org/.../Overfitting

kalo di data mining bisa direduksi pake metode cross-validation (CV n-folds) en, ada baiknya algoritma ybs di uji lagi boss dengan dataset yang lebih besar.

@arifin

text mining juga lebar lho boss.. fungsi text mining nya apakah asosiasi (field information retrieval), sequence (field biogenetik) ? baru bisa ketahuan list algoritmanya.

@adhiguna

boss... NN yang dipakai Norman output-nya true or false...

kalo machine vision pasti outputnya akan lebih dari itu bukan ?

dan melihat kompleksitas dan cost komputasinya..  aku agak ragu jika NN workable untuk machine vision yang perlu response time yang singkat...

# May 22, 2008 8:55 PM

adhiguna said:

@Acoltye, machine vision kalau defect inspection outputnya cuman yes or no kok, cuman kalau masalah response time, memang NN biasanya computation costnya tinggi, so agak lambat, biasanya untuk real application, pakai K-Nearest Neighbor aja udah cukup kok.

Oh ya juga buat Norman, I think I am agree with Acoltye, akurasi 100 persen itu bisa jadi karena overfitting. Coba datanya pake yang lain deh.

# May 25, 2008 8:48 AM