-
I get this as well - either selecting from the drop down box or moving to "next day". I'm using gtd-free-0.5-beta.jar java 1.6.0_11 on windows XP. It's not just a naming issue - it really thinks these events are dated March as they don't appear in the "next month" view.
2009-01-30 19:33:55 UTC in GTD-Free
-
I think you need to also activate the wiimote as a HID. try right-clicking on the wiimote in the BlueSoleil window and looking for something about services.
2008-04-17 16:26:38 UTC in Wiimote Whiteboard
-
I wanted to see how far you'd got with this but the planet-earth domain is not working. Can you post a link to your PhD (when it's published!) and let us know if the site will be back (or the same content can be made available elsewhere).
Thanks
Oliver.
2008-03-25 18:06:36 UTC in The Planet Earth Project
-
I'm no expert but here are a couple of ideas: 1. make sure you are using the latest version of joone (the release candidate from the overnight builds) as many people have reported crashing / hanging of earlier versions; 2. try using the JooneTools methods to run your network (JooneTools.interrogate) to see if there's something about your code that isn't quite right.
2007-12-12 11:54:00 UTC in Java Object Oriented Neural Engine
-
I usually start with the JooneTools.createStandard method, which creates a fully connected network. You can create a custom architecture by changing the FullSynapse between the input and hidden layers: you create a new Matrix object with only the weights you want Enabled (so they learn) and the others set to zero and Fixed. see the setEnabled and setFixed methods for Matrix objects and the...
2007-12-05 10:25:44 UTC in Java Object Oriented Neural Engine
-
It's not clear whether you want to set the bias and weights for the layer or write input values to the layer. The weights and biases are kept in Matrix objects that you can write to directly. If you want to write particular input values to a layer of neurons, you can connect a DirectSynapse to that layer, then write to that. If it's your input layer, the easiest way is to use...
2007-11-18 22:29:42 UTC in Java Object Oriented Neural Engine
-
take a look at the JOONEGAP project on Sourceforge. The version on the download page was based on an earlier version of JGAP, so it doesn't currently work, but I think the version in the CVS contains an update. There's a message on this forum about JOONEGAP 0.5 as well.
Either one will give you ideas about how to create genetic algorithms that work with JOONE. Rather than start from scratch...
2007-08-24 07:44:44 UTC in Java Object Oriented Neural Engine
-
Yes - create your own custom Layer by extending the SimpleLayer class and including your own function in the forward() method. Look at SigmoidLayer or LinearLayer as examples. You also need to consider how you want to adjust weights in the layer based on your bespoke transfer function and implement this in the backward() method.
2007-08-22 10:28:12 UTC in Java Object Oriented Neural Engine
-
You've probably moved on since your post, but I would be interested in temporal difference learning being available in Joone. I think some form of backprop is still used, but the network is trained by comparing its current output to outputs from previous periods. see Chapter 6 of http://www.cs.ualberta.ca/~sutton/book/ebook/the-book.html
Google TD-Gammon for an example of this in use.
2007-08-11 09:08:01 UTC in Java Object Oriented Neural Engine
-
I think you have the same problem as I did. You probably need to get RC2, where this bug has been fixed. See my thread below with an answer from Paolo or look at the other forum "Open Discussion".
2007-08-11 09:01:51 UTC in Java Object Oriented Neural Engine
