When running the simulation, you often see multiple nodes respond to a name ( like two careers) will activate for Sam. Is this because of demographic over lap? Like other members with similar age and status as Sam have that secondary career? So the network draws on that association?
Yes exactly, I believe that's what's happening. Maybe Sam and Joe and Bob (making these names up) are all in the Jets and in their 20's, but they have a few different jobs. Because of these overlaps activating Sam will active the Joe and Bob instance nodes, which will activate the jobs they have. Intuitively, this makes sense. When you hear a friend's name you might think not just about that person's job, but of jobs associated with people _like_ that person. You can see how this could explain things like stereotypes as well.
When I use the one-two method to create my own network, I do not think it is setting up a bidirectional connection. Only on red dot is appearing on one side. What may I be doing incorrectly?
The 1-2 trick does only create one-directional connections. You can get pretty fast at just doing the trick twice to get both connections, but if you want to do it in one go, you can shift-select two neurons, then press 1-2, and it will do what you're looking for.
Hi there. There may be a way to merge these ideas, but in general the purposes are different. SOM's are mainly used in a "feed-forward" way to automatically classify inputs, whereas IAC is a more recurrent way of modeling memory.
I miss the days when people used to settle their differences with choreographed dance routines.
When running the simulation, you often see multiple nodes respond to a name ( like two careers) will activate for Sam. Is this because of demographic over lap? Like other members with similar age and status as Sam have that secondary career? So the network draws on that association?
Yes exactly, I believe that's what's happening. Maybe Sam and Joe and Bob (making these names up) are all in the Jets and in their 20's, but they have a few different jobs. Because of these overlaps activating Sam will active the Joe and Bob instance nodes, which will activate the jobs they have. Intuitively, this makes sense. When you hear a friend's name you might think not just about that person's job, but of jobs associated with people _like_ that person. You can see how this could explain things like stereotypes as well.
When I use the one-two method to create my own network, I do not think it is setting up a bidirectional connection. Only on red dot is appearing on one side. What may I be doing incorrectly?
The 1-2 trick does only create one-directional connections. You can get pretty fast at just doing the trick twice to get both connections, but if you want to do it in one go, you can shift-select two neurons, then press 1-2, and it will do what you're looking for.
Thanks, great presentation. Love the program.
Thanks!
Can I do this with a SOM based neural n/w?
Hi there. There may be a way to merge these ideas, but in general the purposes are different. SOM's are mainly used in a "feed-forward" way to automatically classify inputs, whereas IAC is a more recurrent way of modeling memory.
How do you reset the iterations??
Hi there. Double click on the label that shows the iterations.
@@jeffyoshimi1658 Wow, I feel silly. Thank you so much.
I made one for numbers, high-med-low even-odd, and 2 was kind of even a bit medium and low...
5 is like 7...
8 is exactly the same as 6... hahahha... my creation is dumb... hahahhha
wait... 6 is more 8-like than 8 itself :0
if you don't give it input it assumes the that... i guess... all numbers are like 6 and 8...
thanks
Somebody needs to make a dragon flies mind...