Wow ! The video is in cantonese !
Fortunately, I'm born in hong kong too. But, left the country 35 years
ago. I'm now french speaking. I understood most of what you said but not
all.
Beside that ... I have similar idea than you.
I have a formation in electrical engineering. Neural networks can be
compared to logic circuits in electronics.
A feed forward network works like a combinational logic circuit. For an
input, you have always the same output (if the synapse weights didn't
change in meantime).
A recurrent neural network with its feedback loop works like a sequetial
logic circuit. For the same input, the output may not be the same at
each iteration. In electronics, we consider that it may be equivalent to
a combinational logic circuit with some additional hidden inputs. For
the recurrent neural network, it acts if there are hidden memory.
And the problem is "Is there a way to control this hidden memory ?"
LAU
Le 25/02/2016 08:28, YKY (Yan King Yin, 甄景贤) a écrit :
Main ideas:
1) cognitive states live in a vector space
2) a recurrent neural network acts on the cognitive states
3) reinforcement learning provides top-level control
I'm looking for partners to discuss further / implement the new
prototype :)
*AGI* | Archives <https://www.listbox.com/member/archive/303/=now>
<https://www.listbox.com/member/archive/rss/303/27172223-36de8e6c> |
Modify
<https://www.listbox.com/member/?&;>
Your Subscription [Powered by Listbox] <http://www.listbox.com>
-------------------------------------------
AGI
Archives: https://www.listbox.com/member/archive/303/=now
RSS Feed: https://www.listbox.com/member/archive/rss/303/21088071-f452e424
Modify Your Subscription:
https://www.listbox.com/member/?member_id=21088071&id_secret=21088071-58d57657
Powered by Listbox: http://www.listbox.com
