Comments for Taking Initiative

Comment on DirectX10 Tutorial 9: The Geometry Shader by akismet-a83dca4c2ab5a8a4489ff45d7efe8a1f

akismet-a83dca4c2ab5a8a4489ff45d7efe8a1f — Tue, 15 May 2012 15:47:05 +0000

Hi Bobby.
Thanks for this tutorial, it rely help me out allot.
Any idea when the next tow parts are coming?

Comment on Basic Neural Network Tutorial – Theory by samira

samira — Tue, 08 May 2012 16:53:59 +0000

i love u so nice topic i enjoyed alot

Comment on DirectX10 Tutorials by Shen Weixin

Shen Weixin — Tue, 08 May 2012 02:45:37 +0000

http://www.rastertek.com/
this’s a quite good DX tutorial website.

Comment on C++ and XML: what a nightmare by WSemmelink

WSemmelink — Fri, 04 May 2012 13:37:30 +0000

It looks like tinyxml got a face lift: TinyXML2
http://www.grinninglizard.com/tinyxml2/index.html
Still no SAX, XSLT, XPATH etc.

Comment on Basic Neural Network Tutorial : C++ Implementation and Source Code by pathwaystoknowledge

pathwaystoknowledge — Thu, 03 May 2012 19:30:37 +0000

Ah! slowly the light bulb starts to go on.

I tried setting max epoch to 250, desired accuracy to 98, then looping by 4′s over number of input and hidden nodes. output is #input nodes, #hidden nodes, epochs to termination

Data File Read Complete >> Patterns Loaded: 20000
4,4,250
4,8,250
4,12,250
4,16,250
4,20,250
4,24,250
8,4,250
8,8,250
8,12,250
8,16,250
8,20,250
8,24,250
12,4,250
12,8,250
12,12,250
12,16,250
12,20,250
12,24,250
16,4,250
16,8,250
16,12,250
16,16,250
16,20,250
16,24,250
20,4,2
20,8,2
20,12,2
20,16,2
20,20,2
20,24,2
24,4,2
24,8,2
24,12,2
24,16,2
24,20,2
24,24,2
– END OF PROGRAM –

(this, incidentally, takes a really freaking long time – 90 minutes on my machine)

Nothing like data 8cD

Comment on Basic Neural Network Tutorial : C++ Implementation and Source Code by pathwaystoknowledge

pathwaystoknowledge — Thu, 03 May 2012 18:14:41 +0000

obviously I’ve also changed the hidden layer to 16 here, but behavior’s the same with hidden layer at 8

Comment on Basic Neural Network Tutorial : C++ Implementation and Source Code by pathwaystoknowledge

pathwaystoknowledge — Thu, 03 May 2012 18:10:43 +0000

How odd-
I was messing about with the code, and changed the number of input nodes to 24 just out of monkey curiousity. Sure didn’t expect:

Data File Read Complete >> Patterns Loaded: 20000

Neural Network Training Starting:
==========================================================================
LR: 0.01, Momentum: 0.8, Max Epochs: 500
24 Input Neurons, 16 Hidden Neurons, 1 Output Neurons
==========================================================================

Epoch :0 TSet Acc:5.225%, MSE: 0.142681 GSet Acc:10.025%, MSE: 0.101812
Epoch :1 TSet Acc:58.525%, MSE: 0.0364501 GSet Acc:91.35%, MSE: 0.00485578
Epoch :2 TSet Acc:96.4583%, MSE: 0.00301354 GSet Acc:99.475%, MSE: 0.00234604
Epoch :3 TSet Acc:99.6333%, MSE: 0.00163814 GSet Acc:99.8%, MSE: 0.00170634

Training Complete!!! – > Elapsed Epochs: 500
Validation Set Accuracy: 99.975
Validation Set MSE: 6.97053e-005

– END OF PROGRAM –

total changes from d/l:
changed line 25 of main.cpp to
neuralNetwork nn(24, 16, 1);

sort of expected it’d blow up or perform identically to the install settings.

Comment on DirectX10 Tutorial 9: The Geometry Shader by Bobby

Bobby — Tue, 01 May 2012 13:59:52 +0000

you dont actually map a texture to the screen, you are mapping the texture onto a quad that you create in the GS. So it’s exactly the same as if you’d have rendered a quad and textured it. The only trick is to orient the quad so that its always facing the camera or in the same position relative to the camera (e.g. when rendering GUIs).

Comment on Basic Neural Network Tutorial : C++ Implementation and Source Code by mario

mario — Sun, 29 Apr 2012 11:58:50 +0000

I’ve trained the NN on my datasets and now I want to use the weights of the trained NN on testing on other user given data? (I’m using NN to do regression not clssification, so is you model suitable for regression as well?)

Comment on DirectX10 Tutorial 9: The Geometry Shader by Michael

Michael — Wed, 25 Apr 2012 12:30:48 +0000

Hi, thanks for another great tutorial . I was wondering if you could maybe explain a bit more how the transformation happens onto screen space ( I don’t really need it right now, but its driving me crazy that I don’t understand whats going on ).

What exactly happens when you are trying to map a certain texture to a screen at all times ? It sounds more logical for me to map entire 3D objects in a world than it is to have a simple image on my screen .

thanks