Yes but I won't be clustering vectors nor using backprop. I will however later 
store dog related words under a node though, maybe, which uses a cluster to 
quickly learn to relate new words.

Is the dictionary I use from cmix (which turns enwik8.txt from 100MBs to 61MBs) 
grouping related words? From what I can know, it just gives words shorter 
codes, however looking at some tests below, it seems to often give related 
words similar codes? This can't be good if so, it is not considering 
probabilities...I wonder how much the smaller codes and how much the grouping 
is helping my compression?



and and and and then then then then mom when dad and and and and then then then 
then mother when father and and and and then then then then dog when cat

        Ô¤ Ô¤ Ô¤ Ô¤ mom ½ dad         Ô¤ Ô¤ Ô¤ Ô¤ ê  ½ ê¡         Ô¤ Ô¤ Ô¤ Ô¤ 
ò¹ ½ ò¸



after after after after mom after after after after dad after after after after

after after after after mother after after after after father after after after 
after

after after after after dog after after after after cat after after after after 
horse after after after after pig after after after after man after after after 
after

after after after after food after after after after meal after after after 
after lunch after after after after dinner after after after after

after after after after walk after after after after run after after after 
after jog after after after after move after after after after

¾ ¾ ¾ ¾ mom ¾ ¾ ¾ ¾ dad ¾ ¾ ¾ ¾

¾ ¾ ¾ ¾ ê  ¾ ¾ ¾ ¾ ê¡ ¾ ¾ ¾ ¾

¾ ¾ ¾ ¾ ò¹ ¾ ¾ ¾ ¾ ò¸ ¾ ¾ ¾ ¾ ò° ¾ ¾ ¾ ¾ pig ¾ ¾ ¾ ¾ ê± ¾ ¾ ¾ ¾

¾ ¾ ¾ ¾ æ¡ ¾ ¾ ¾ ¾ ÷ì› ¾ ¾ ¾ ¾ ÷é€ ¾ ¾ ¾ ¾ öÔÀ ¾ ¾ ¾ ¾

¾ ¾ ¾ ¾ ÑŒ ¾ ¾ ¾ ¾ ÐÉ ¾ ¾ ¾ ¾ jog ¾ ¾ ¾ ¾ а ¾ ¾ ¾ ¾
------------------------------------------
Artificial General Intelligence List: AGI
Permalink: 
https://agi.topicbox.com/groups/agi/T192296c5c5a27230-M7d50af85c3e90222e372d498
Delivery options: https://agi.topicbox.com/groups/agi/subscription

Reply via email to