Wednesday, September 30, 2009

CCK09: There's too much salt in my conceptual network

I think there are two self-organizing learning networks. One is the neural network and the other is the bit network. The human and the computer. The beauty of connectivism is its explanation of how these two networks connect such that two humans can share neural networks.

Unfortunately humans do not directly connect neurons, nor do we connect neurons to bits. Our communication with other humans and with computers is mediated by language. And imho language is an arbitrary system of symbols i.e. words, gestures, voice, and images.

It is (unfortunate that) these symbols which make up our conceptual network. Even if one where to deny a symbolic system in connectivism, he/she would still be forced to use language to deny that. I think it would be more practical to admit the limitations of our languages and try to approximate the "if" neurons of one person can directly communicate with the neurons of another.

What had bothered me about the current state of connectivist explanation of concepts is the selection of what nodes should represent. In the neural level it is explicit. In the social level it is natural to identify an actor as the node. The human is after all a world in himself/herself. But in the conceptual level, made up of words, it is not so clear. I think a lot of what is represented as nodes should be represented as ego networks or component networks.

This could be illustrated in Chemistry. If we consider salt for instance. Salt is not represented as a node but a network of basically Sodium and Chlorine, connected by bonds (please correct me if I'm wrong because I'm not a chemist). I think we misrepresent a lot of concepts like salt as nodes in our conceptual layers, and have not gotten around to identifying the basic elements of a learning conceptual network that are simple enough to have no meaning when standing by themselves. Resulting in a salty network.

Sunday, September 6, 2009

Stopmotion animation with a webcam in UbuntuStudio

This educational tool is probably for under K-6 education. I still can't think of an example for secondary and higher ed students.

The Stopmotion software for stop motion animation is included in the distribution of UbuntuStudio. The demo here was done with UbuntuStudio 8.04 based on Intrepid.

I used an A4Tech web cam which is v4l2. Stopmotion has a problem with these kind of web cameras so you'll need to follow Aearenda's solution here: http://ubuntuforums.org/showthread.php?t=931208 . I do hope they've already solved this in Ubuntu Jaunty 9.04.

Create an mplayer video import device in Stopmotion


Start Stopmotion and in the menu find settings->configure stopmotion or ctrl-P.

Click add button.

Resize the dialog box to see the start deamon & stop deamon text boxes.

Click video import tab.

In the start deamon text box enter the following (change the ~/RENDER/stopmotion/ folder with your preference):

~/RENDER/stopmotion/startmplayer $VIDEODEVICE $IMAGEFILE 2 0.25s .tempjpgs &


In the stop deamon text box enter this code:

killall startmplayer && killall mplayer


Download my startmplayer file from: http://www.mediafire.com/file/zm2ytmnqa2j/startmplayer.zip.

Extract this startmplayer file in the ~/RENDER/stopmotion/ folder, you can change this folder.

Then follow this demo:



Finished product (a bit crude):

Saturday, September 5, 2009

Symbolic versus Connectionist Simulation

While happily trying to learn Soar due to my quest for my own simulated student, I came across Allen Newell's statement that:

"The final risk [to unified theories of cognition] is the rising tide of connectionism, which is showing signs of sweeping over all cognitive science of the moment. The excitement is palpable--we are all hot on the trail of whether neuroscience and th cognitive world can finally be brought together. That is indeed an exciting prospect. But my message relates to symbolic architectures and all the other good things that connectionism sees as the conceptual frame to overthrow. So the final risk is that my timing is terrible. John Anderson had it right when he wrote his book on the architecture of human cognition for publication in 1983. It is too late now. (1994)"

My heart sank when I realized I wasn't paying attention to the issue of symbolic cognitive science and connectionist cognitive science. The fact is that connectivism appear to be siding with connectionist. And here I am fiddling with a symbolic cognitive architecture, Soar. I had to look for a connectivist i.e. neural network simulation tool, and found emergent. The question is whether emergent can be used to build a simulated student. It looks so daunting, and appears to be rooted in computational simulations.

Reference:

Newell, A. (1994). Unified theories of cognition. USA: Harvard University.
 
Creative Commons License
This work is licensed under a Creative Commons Attribution-Share Alike 3.0 Unported License.