Java Neural Network Framework Neuroph Activity
Brought to you by:
sevarac
Activity for Java Neural Network Framework Neuroph
-
Sean O'Connor posted a comment on discussion Open Discussion
SwitchNet in Java: SwitchNet
-
SwitchNet, a fast low parameter count neural network:: https://editor.p5js.org/siobhan.491/sketches/RvqZfikaE Blog: https://ai462qqq.blogspot.com/2023/04/switch-net.html In the blog comments there is a near Java version (ie. processing).
-
You can 2 side ReLU via its forward weight connections (not its backward weight connects). Hence you are involving the next layer in things. Blog post: [2 Siding ReLU]https://ai462qqq.blogspot.com/2023/03/2-siding-relu-via-forward-projections.html/) Also there is a way of streamlining a neural networks layer. Allowing you to reasonably use very wide layers. Blog post:Reduced cost layers
-
Duane Kunkee posted a comment on discussion Help
I am getting the same thing on Windows 11. A message indicates that I should provide a javahome parameter, but there doesn't seem to be a way to do it. I tried defining a javahome environment variable, but that didn't make any difference. Is there anyone seeing this that has successfully installed Neuroph on Windows 10 or 11?
-
Radoslavov posted a comment on discussion Help
Hi! Your training set is empty.
-
Gianni Casonato posted a comment on discussion Help
Hi, I have some trained networks generated using other libraries, and I wonder if Neuroph is able to handle some sort of standard exchange formats (.zip in particular including weights in hdf5 or bin, and model in yaml or jason). From the examples I've only seen load of .nnet files generated by Neuroph itself. Thanks in advance and regards, Gianni
-
Thiemo posted a comment on discussion Help
Hey Guys, can you help me a little bit out with neuroph. I am new in neural networks. I have a lot of background knowledge, such as types, history usw of neural networks but i fail to teach a network anything. I already tried a few NeuroTypes and already played with the neuroph studio, where the training usw always works :( So heres my current code, very compact its a "how to do ki" learning project xD NeuralNetwork<BackPropagation> neuralNetwork = new MultiLayerPerceptron(2, 12, 1); BackPropagation...
-
Pieter Svenson posted a comment on discussion Help
Hello! I have an issue with creating a custom neural network. I detailed out my issue on StackOverflow. Can someone take a look and try to find the problem? https://stackoverflow.com/questions/69932361/neuroph-api-custom-network-not-configured-correctly-outputs-0
-
JakobB posted a comment on discussion Help
Hi, I trained a CNN with MNIST, following the example in the github-repository. However the results are always around 10% accuracy, even after 100 epochs. Is there a known problem in the source with v2.98?
-
Peter Lin posted a comment on discussion Help
Here is the error from the IDE log when I click "new project" SEVERE [global] java.lang.ClassNotFoundException: javax.activation.DataContentHandlerFactory at java.base/jdk.internal.loader.BuiltinClassLoader.loadClass(BuiltinClassLoader.java:606) at java.base/jdk.internal.loader.ClassLoaders$AppClassLoader.loadClass(ClassLoaders.java:168) at java.base/java.lang.ClassLoader.loadClass(ClassLoader.java:522) at org.netbeans.ProxyClassLoader.loadClass(ProxyClassLoader.java:222) Caused: java.lang.ClassNotFoundException:...
-
Zoran Sevarac posted a comment on discussion Help
Thank you! It's always best when you dig a little bit and find solution for yourself. That's the best way to learn.
-
Thanks for the Pack200 fix. Can you provide more info how the new project fails? Thanks
-
It seems like there is a bug in training workflow. Will put it on todo to fix for next release. Running training with only basic options should work
-
Given M1 Macbook air defaults to using zsh now, I was able to download the zip version and get it to run. But after it starts, creating a new project fails. here's the steps to get it to run open terminal run "/bin/bash" run neuroph studio with ./neurophstudio --jdkhome [path to jdk home] in my case, I am running Azul zulu openjdk release for M1. The path zulu uses is /Library/Java/JavaVirtualMachines/zulu-15.jdk/Contents/Home/
-
Exception: java.lang.NoClassDefFoundError. java/util/jar/Pack200 I'm using Azul's openJDK build for M1 mac. openjdk version "15.0.1" 2020-10-20 OpenJDK Runtime Environment Zulu15.28+1013-CA (build 15.0.1+9)
-
I downloaded the zip version of release 2.98 and followed the docs for creating a basic test project. http://neuroph.sourceforge.net/image_recognition.html When I try to run the training wizard, it gives me a error no matter what value I use for Training set percent SEVERE [com.jme3.gde.core.scene.SceneApplication]: Uncaught exception thrown in Thread[Thread-13,6,main] java.lang.ClassCastException: org.neuroph.training.Variable cannot be cast to java.util.Stack at org.neuroph.training.SetWorkflowParametarsTask.execute(SetWorkflowParametarsTask.java:29)...
-
Hello! Please if you can answer? I am having a Perceptron with one input and one output. Since i found no update and no output for so long. Just to shrink the problem i do with one by one. I am creating layers, adding single neurons, setting properties, attaching to output and input and ussing LayerFactory to connect them. I don't know where i do mistake but finnaly i get the next: Input Connection between Layer 1, the only neuron - 0 AND Layer 0, the only neuron - 0 [Connection{fromNeuron=Neuron{totalInput=0.1111111111111111,...
-
Hello! Please if you can answer? I am having a Perceptron with one input and one output. Since i found no update and no output for so long. Just to shrink the problem i do with one by one. I am creating layers, adding single neurons, setting properties, attaching to output and input and ussing LayerFactory to connect them. I don't know where i do mistake but finnaly i get the next: Input Connection between Layer 1, the only neuron - 0 AND Layer 0, the only neuron - 0 [Connection{fromNeuron=Neuron{totalInput=0.1111111111111111,...
-
Hello! Please if you can answer? I am having a Perceptron with one input and one output. Since i found no update and no output for so long. Just to shrink the problem i do with one by one. I am creating layers, adding single neurons, setting properties, attaching to output and input and ussing LayerFactory to connect them. I don't know where i do mistake but finnaly i get the next: Input Connection between Layer 1, the only neuron - 0 AND Layer 0, the only neuron - 0 [Connection{fromNeuron=Neuron{totalInput=0.1111111111111111,...
-
Hello! Please if you can answer? I am having a Perceptron with one input and one output. Since i found no update and no output for so long. Just to shrink the problem i am creating layers, adding single neurons, setting properties, attaching to output and input and ussing LayerFactory to connect them. I don't know where i do mistake but finnaly i get the next: Input Connection between Layer 1, the only neuron - 0 AND Layer 0, the only neuron - 0 [Connection{fromNeuron=Neuron{totalInput=0.1111111111111111,...
-
Rucelle C. Garciano posted a comment on discussion Help
Hi! I need help. I've just currently downloaded and installed Java JDK in my laptop with WIndows 10 Pro OS x64 processsor. But when I would try to install Neuroph Studio afterwards, it asks for Java JDK to be installed well in fact, it has already been installed... How must I proceed to solve this problem?
-
Joao Madeira posted a comment on discussion Help
Also tried the zip and I can't create projects.
-
same here. Even with the last 2 versions
-
Mr. G posted a comment on discussion Help
Same here...
-
Gareth Evans posted a comment on discussion Help
I'm attempting to install NeuroPH 2.98 on a windows 10 PC. It fails to start the installer (neurophstudio-windows-2.98.exe) and gives me the error "Java SE Development kit (JDK) was not found on this computer" I have installed JDK 1.7.0_80 just for this system, into the default location. Any advice?
-
Yasir Ali posted a comment on discussion Help
share your whatsapp number as I can discuss with you through voice chat On Monday, February 1, 2021, Yasir Ali yasirali988@users.sourceforge.net wrote: How may I help you ? https://www.avast.com/sig-email?utm_medium=email&utm_ source=link&utm_campaign=sig-email&utm_content=webmail&utm_term=icon Virus-free. www.avast.com https://www.avast.com/sig-email?utm_medium=email&utm_ source=link&utm_campaign=sig-email&utm_content=webmail&utm_term=link <#DAB4FAD8-2DD7-40BB-A1B8-4E2AA1F9FDF2> On Wed, Jan 27,...
-
How may I help you ? https://www.avast.com/sig-email?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail&utm_term=icon Virus-free. www.avast.com https://www.avast.com/sig-email?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail&utm_term=link <#DAB4FAD8-2DD7-40BB-A1B8-4E2AA1F9FDF2> On Wed, Jan 27, 2021 at 6:48 AM Edgardo Vazquez edgardo369@users.sourceforge.net wrote: I'm new to Neuroph too. Did you get your question answered? How to extract features...
-
Edgardo Vazquez posted a comment on discussion Help
I am trying to do a project to analyze ~1000 data points. Each data point is composed of 15 independent numerical variables and one dependent numerical variable. I would like a deep learning analytics platform to analyze this data and return to me a model of some kind where I can input values for a dependent variable and have the model/program return to me the values that the 15 independent variables would need to have in order to hit the provided dependent variable value (I'd also like the program...
-
I'm new to Neuroph too. Did you get your question answered?
-
hellp everyone Please Help me and tell me How neuroph studio Calculates accuracy of any problem, and what is the functionality of Neuroph studio ? Means How it works how it process the data behind the seen .
-
hello everyone , Iam new to Neuroph studio I wish to do a project on disease diagnosis my teacher asked me how will you extract the features and which classifier will use for your project and how will you selecf the data, Please Guide me about these questions . I took data from different patients I will train the data please help me solve my question .
-
Georg Lehnhardt posted a comment on discussion Help
I'm using neuroph 2.92 and created an instance of MLP with the e.g. following topology: 50(input), 101 (2* input + 1), 1(output); TransferFunctionType = SIGMOID and the MomentumBackpropagation has following parameters:MAX_NUMBER_OF_ITERATIONS = 10000; LEARNING_RATE = 0.5; MOMENTUM = 0.7; MAX_ERROR = 0.0001. Durring the training the network error is smaller the MAX_ERROR, durring validation I obtain a MSE - value between 0.24 and 0.9, wich is OK , but durring the test - phase, I obtain with theese...
-
Aleh posted a comment on discussion Help
Hi, starting from Tuesday neuroph.sourceforge.net/maven2/ is no more available under HTTPS It works with HTTP but it is not an option Could you please help?
-
Liuri You posted a comment on discussion Help
on the Web page http://neuroph.sourceforge.net/tutorials/MultiLayerPerceptron.html Q1. On the dialogue of parameters, we specify 3 hidden neurons and don't specify the number of hidden layers, why would the created network have 2 hidden layers with 3 and 4 neurons respectively, and 7 hidden neurons in total? Q2. Also on that dialogue, we let the Transfer function be the default sigmoid, why it's said that "All neurons will have Tanh transfer functions" after the creation of the network? Q3. At the...
-
Java Neural Network Framework Neuroph released /neuroph-2.98/RELEASE_NOTES.txt
-
-
-
Helmut Wiedemann posted a comment on discussion Help
I istalled Neuroph Studio on opensuse leap 15.1 with the shell-script. It starts normally, but I cant use some important buttons, like "New Project" and many others: Nothing happens! I am using GNOME. Maybe it has something to do with JavaFX, which is not activated in the plug-in-section? My Java Version on my system is 11, but i can downgrade to Java 8 - but this has no effect. I also tried to switch to KDE, mybe GNOME is the problem. But: The same thing! I think, there must be a problem with a...
-
Colors correspond to neuron activation Red == 1 Greay == 0 Blue == -1 You can turn on display preferences to show values, and you can incpect indicvidual neurons in properties
-
James M. Yunker posted a comment on discussion Help
I am confused about the colors of the neurons on neuroph. It seems that most of the time they are pale gray, but sometimes red, blue and green. I do not know what is the meaning behind the colors and what does it mean when they change colors? Any help appreciated. Thanks in advance. Respectfully, Lou Reed
-
Sam Cleveland posted a comment on discussion Help
I wrote a neural network which it's goal was to be able to calculate the modulus of two 8-bit binary numbers. There is 16 input neurons, 8 output neurons and two layers of 32 hidden neurons. The code is as follows: import java.util.Arrays; import org.neuroph.core.NeuralNetwork; import org.neuroph.core.Neuron; import org.neuroph.core.data.DataSet; import org.neuroph.core.data.DataSetRow; import org.neuroph.core.learning.IterativeLearning; import org.neuroph.core.learning.LearningRule; import org.neuroph.core.learning.SupervisedLearning;...
-
Valentin posted a comment on discussion Help
Hi ! Plz can someone told (or show turtorial) how to integrate compleated NN (.nnet) to eclipse ? Step by step. I am not java programmer so i need very detailed instruction (from simple how and what lib need to be inserted) to final code line where it is possible to insert values in NN and get result (i understand how in java make simple manipulations and i will be able to set up variables from sorse)
-
RALPH ALBERT SOMMER posted a comment on discussion Help
Install jdk 8 and add the variable jdkhome=<path> in neurophstudio.conf.</path>
-
hello Seniors and experts I am researching on typhoid fever recognition using neural network I collected data set of 100 patients now my question is that how to give input in data set as trainning can be done easily , for example if a patient have fever positive so how to give input of fever in data set ?? help me to train data which I have collected .
-
hello Seniors and experts I am researching on typhoid fever recognition using neural network I collected data set of 100 patients now my question is that how to give input in data set as trainning can be sone easily , for example if a patient have fever positive so how to give input of fever in data set ?? help me to train data which I have collected .
-
Anthony Olevano posted a comment on discussion Help
This is my code: public class Test { private static NeuralNetwork ann = null; public static void main(String args[]) { ann = assembleNeuralNetwork(); trainNeuralNetwork(ann); ann.setInput(0, 1); ann.calculate(); print("0, 1", ann.getOutput()[0], 1.0); ann.setInput(1, 0); ann.calculate(); print("1, 0", ann.getOutput()[0], 1.0); ann.setInput(0, 0); ann.calculate(); print("0, 0", ann.getOutput()[0], 0.0); ann.setInput(1, 1); ann.calculate(); print("1, 1", ann.getOutput()[0], 0.0); } private static void...
-
Yannick posted a comment on discussion Help
Hi, I would like to know if it is possible to create and test recurrent neural networks with NeurophStudio ? BR
-
Hi, I would like to know if it is possible to create and test recurrent neural networks with NeurophStudio ? BR
-
Ramon posted a comment on discussion Open Discussion
I'd like to create a simple ANN to predict the price of gold going up or down based on the price of silver. I'd like to create a training set, but I'm unsure how to do this as DataSet can only contain double values, and i'd assume that I'd need a time factor. Any help would be much appreciated. Currently, I'm assuming that I'd need a series of: DataSetRow rOne = new DataSetRow(new double[] {priceOfSilver, time}, new double[] {priceOfGold}); to do this, but I'm unsure. Thanks. Help is much apprec...
-
Sean O'Connor posted a comment on discussion Open Discussion
I guess current artificial neural networks have inherited a lot of bagage from the 1950s onward. I think the foundations have just been taken on trust and never really looked at again or checked. Anyway: https://discourse.processing.org/t/flaw-in-current-neural-networks/11512
-
-
-
-
-
-
-
-
-
Joe Salmon posted a comment on discussion Help
This is the error message I get joe@joe-Latitude-E7240:~$ sudo /bin/sh "/usr/local/neurophstudio/bin/neurophstudio" WARNING: An illegal reflective access operation has occurred WARNING: Illegal reflective access by org.netbeans.ProxyURLStreamHandlerFactory (file:/usr/local/neurophstudio/platform/lib/boot.jar) to field java.net.URL.handler WARNING: Please consider reporting this to the maintainers of org.netbeans.ProxyURLStreamHandlerFactory WARNING: Use --illegal-access=warn to enable warnings of...
-
I'm totally new to this and barely know any java and only some very basic stats I remember from uni. I've managed to write a simple hangman game which guesses 10 random letters and I've have created data sets which take the 10 letters guessed as a number from 0-25 as the inputs then if one of the 5 possible words was guessed correctly as either 0 for not guessed or 1 for guessed correctly as output. I've then normalise the data using the max min feature and I've been messing about trying to train...
-
Lax Lam modified a comment on discussion Open Discussion
-
fshuk1 posted a comment on discussion Help
I am trying to port a program from Matlab to java. Matlab was used because it has very comprehensive neural network implementation. I want now to move the project to Java. I am looking for a comprehensive library in Java and have come across Neuroph. So to start with I need to run a very simple example to make sure all is working before trying to port everything. I came across this tutorial. https://www.baeldung.com/neuroph. I tried to implement it in Eclipse. The implementation has no errors with...
-
Hi, I am using the neuroph library for training. Firstly, I normalize my training data into the interval [0 1], using the sigmoid function and backpropagation. After trained the network, I used the testing data to do testing (also in 0-1), the outputs from the trained network are obviously also within [0 1]. But, how can I get the output data in actual data range, not 0-1? Any help would be appreciated.
-
Hi, after several day's frustration, I finally gave up neuroph studio, instead, using neuroph library on eclipse directly, and it turns out working well. You may also just directly writting the code.
-
Hi, I am using the neuroph library for training. Firstly, I normalize my training data into the interval [0 1], using the sigmoid function and backpropagation. After trained the network, I used the testing data to do testing (also in 0-1), the outputs from the trained network are obviously also within [0 1]. But, how can I get the output data in actual data range, not 0-1? Any help would be appreciated.
-
Julio Buonfigli posted a comment on discussion Help
Same problem... I tried older versions but nothing...Any help?
-
Hello, I just use adaptive learning rate for my NN, which is one of the greatest feature (Dynamic Backpropagation) provided by Neuroph. However, when I choose 'Dynamic Backpropagation' for my NN, and press 'train' button, a 'Backpropagation Setting' is shown; I check 'use dynamic learning rate' and press 'train' button, nothing happened, the 'Backpropagation Setting' is still there. Why? any help would be appreciated.
-
Hello, I would like to use adaptive learning rate for my NN, which is one of the greatest feature (Dynamic Backpropagation) provided by Neuroph. However, when I choose 'Dynamic Backpropagation' for my NN, and press 'train' button, a 'Backpropagation Setting' is shown; I check 'use dynamic learning rate' and press 'train' button, nothing happened, the 'Backpropagation Setting' is still there. Why? any help would be appreciated.
-
The version of Neuroph that I use is 2.94(the latest version). When I displayed the Total Network Error Graph, the 'Iteration' at the bottom of the graph is decimal form, such as 1.0 1.5 2.0 2.5 3.0... what are these numbers? and how to get the real iteration number?
-
Hi, I am using the latest version of Neuroph studio (2.94). I find that a palette is always displayed on the right hand side of the interface, quite a lot things (layers, neurons, inut functions...) are shown. But when I click it (or double click...), no any response happened, so I wonder if this palette is just a kind of decoration? any help would be appreciated.
-
Thank you. However, what if I want to display the learning curve of both the training and testing. Neuroph will display an Error Graph (Is this the learning curve?) after click the 'train' button. But if I want the learning curve of the testing and I can't click the 'train' button, how can I get the learning curve?
-
Thats the right way, only when you drag test set dont clcik train, but test button in toolbar And if you click only test it wont destroy your allready trained network just make sure you've saved it before
-
Hi thanks for the note. It seems that here is some kind of misconfiguration on source forge host so many links ar ebroken, investigating... We're thinking of moving all tutorials to git hub so they could be updated by the community. In the meantime check out the code samples there is also wine example in Java code https://github.com/neuroph/neuroph/blob/master/neuroph-2.9/Samples/src/main/java/org/neuroph/samples/uci/WineClassificationSample.java Cheers
-
Hi, I'm new to neural network, and Neuroph Studio could be a very usefull toolkit for my project, so really thank you for the development team. But I've encounted a quite annoyed problem about the toolkit. I am using the lastest verrsion of neuroph studio. After I dragged my training set to the input of the network(MLP) and trained, I want to test my network using my testing set, but do I have to drag the testing set to the network inputs again and click 'train/test' button, won't that destroy my...
-
David Johnson modified a comment on discussion Help
Hi, I am trying to learn a neural networking technology in java, and neuroph was highly recommended. I am trying to follow the tutorials (Wine and Glass classification), but the instructions seem to be geared to an older version of the software. Menus and UI have changed, so it's hard to follow the tutorials. When I navigate the site to download the sample training sets, I get the error (below) from the server. Is there another way to pursue these tutorials? The tutorials talk about normalizing the...
-
Hi, I am trying to learn a neural networking technology in java, and neuroph was highly recommended. I am trying to follow the tutorials (Wine and Glass classification), but the instructions seem to be geared to an older version of the software. Menus and UI have changed, so it's hard to follow the tutorials. When I navigate the site to download the sample training sets, I get the error (below) from the server. Is there another way to pursue these tutorials? The tutorials talk about normalizing the...
-
Hi, I am trying to learn a neural networking technology in java, and neuroph was highly recommended. I am trying to follow the tutorials (Wine and Glass classification), but the instructions seem to be geared to an older version of the software. Menus and UI have changed, so it's hard to follow the tutorials. When I navigate the site to download the sample training sets, I get the error (below) from the server. Is there another way to pursue these tutorials? The tutorials talk about normalizing the...
-
Ezio Thien posted a comment on discussion Help
Wait, first of all how did you use the image recognition in netbeans?
-
Cleber Amaral posted a comment on discussion Help
In fact, I have also experienced the same error. The workaround was to remove the first line of my csv, the row of columns names.
-
starf posted a comment on discussion Open Discussion
Just uncomment line 141 of "org.neuroph.nnet.learning.ResilientPropagation", it'll correct the problem. //weight.value += weightChange; -- ovo mora da se radi simultano => weight.value += weightChange; //-- ovo mora da se radi simultano
-
Lawrence modified a comment on discussion Help
I recently got the image recognition to work with netbeans and its printing out both the outputs but the output line is a bit to long is their anyway to lower it down for 5 numbers instead per output? also if thats possible is their a way to only get the output which is over 0.7 to display and ignore the otheroutput if its below that ammount?
-
I recently got the image recognition to work with netbeans and its printing out both the outputs but the output line is a bit to long is their anyway to lower it down for 5 numbers instead per output? also if thats possible is their a way to only get the output which is over 0.7 to display to cause a action to happen?
-
arman alam posted a comment on discussion Open Discussion
hi! i am working on an image recognition project. i have used neuroph library in netbeans and successfully built the project. all thanks to Neuroph Studio team. But working on a large number of images lets say 10000 images takes a lot of time to train the network. Each time i add a new image i have to again create and train the network with 10000+1 images. its a bit and time consuming. so my question, is it possible to train a neural network which is already been trained previously with a large number...
-
<REDACTED> posted a comment on discussion Help
Hello..I want to knaw is it possible to create a MLP with diffenert layers have defferent activation functions in neuroph? for an example 1st layer contains 5 nurons with log-sigmoid activation function and next layer have 5 neurons with tan-sigmoid activation function?
-
Jon Barrilleaux posted a comment on discussion Help
is there a way to single step through rows in a test data set to see how each is handled in the the net?
-
To help others new to this, the "trick" is to use the "OK" button, not the "close" button, even though they both offer to save the data. In spite of what seems reasonable, all other attempts to save the data (close then save, save via the main menu) will not work. (If this is SOP for NetBeans I think there are much better and less deceptive GUI conventions).
-
I entered it by hand, hit the "close" button, the system asked if I wanted to save it, I hit the "save" button. When I reopen the dataset it appears empty. I also tried using the "save" menu entry, but with the same result. Any suggestions? What sequence do you use that is known to work?
-
What is the output? Win / lose? Dont normalize as continious value since it doesent ake sense for this domain. Use one input neuron for each card, and binary input vector. Let us know if you get some interesting results
-
Did you entered the data or imported from csv file? You shoul dbe able to use th edat aset right away...
-
You cannot open it in netbeans, but you can open it in Neuroph Studio
-
Thanks for the note Ken. We made some changes in 2.94 in order to make it easier to follow mathematical models of more backprop variations, but it seems that resilient is now broken. It will be fixed with next release.
-
Jose Juan Duarte posted a comment on discussion Help
I have a neural network made in neuroph and now I want to manipulate it in netbeans but I do not know how to do it I know I need an .nnet file but I can not know where to locate it and the libraries do not recognize them as netbeans. If my question was already answered ... please tell me where to find the answer Thank you very much.
-
Ken Hubacher posted a comment on discussion Open Discussion
2.9.2 Works as well. So it would appear the problem occurred after 2.9.2.
-
I've been using Neuroph for several years and I just noticed that Multiperceptron in 2.9.4 doesn't appear to work properly. I created a very basic test 2 inputs, 1 output 1, -1, 1 -1, -1, -1 1, 1, 1 -1, 1, 1 Multiperceptron, 2 input, 2 hidden, 1 output, TANH, Resilient backprop, bias node Learning rate .3, max error .0001, up to 200,000 iterations (a bit overkill) This should learn very easily and very well. If I substitute the 2.8 neuroph core jar and run this test 20 times repeatedly I get the...
-
I'm new to Neuroph and netbeans. Using the latest version (NeurophStudio 201609300101). Folowing a getting started tutorial and video I created a neural net, then tried to create a simple data set. try as I might I could not save a data set for use in training -- it always ends up empty. I tried several ways, and even tried reinstalling the app but with the same result. Is there something obvious I'm missing? Is there something unique about netbeans for creating and saving data/files?
-
Jeff posted a comment on discussion Open Discussion
I'm trying to create a neural network that will bid pinochle hands. To this end, I had my game play itself thousands of times and store the hand and the result. Pinochle deck consists of two each of A, 10, K, Q, J, 9 I tried storing my hand like this: 2,1,0,0,1,1,0,0,1,1,1,0,1,1,0,0,0,0,0,0,47 Where the first 20 spaces represent how many of each card it had (either 2, 1, 0) I trained a MultiLayerPerceptron with 20 inputs and 1 output and the results were nowhere close to as good as my own AI creation,...
-
Dave Dyer modified a comment on discussion Help
All my training examples are unique triplets. Effectively i have an infinitely large training set. So I don't see how over fitting can be relevant.
-
no, but if the transfer function has a finite domain, and the backpropagation algorithm doesn't take that into account, the theory is at least incomplete.
-
Relating to the other network training problems mentioned in my recent posts, I found two other problems in the backpropagation algorithm used in neuroph. The first is the way error is propagated to hidden layers. The sum of the individual errors of the output layer (times the learning rate) is applied to each neuron in the hidden layer. Since typically, all the neurons in the hidden layer see the same sum, the amount of error propagated depends on the number of neurons. For example if the hidden...
1 >
