[FRIAM] On old question

uǝlƃ ☣ gepropella at gmail.com
Fri Oct 26 14:17:16 EDT 2018


Hm.  I'm probably a victim of my own skimming.  Allowing metaphor to run rampant ... The scaffolding Nick linked to is definitely *supervised* in what seems a fairly biased (maybe in a good way) constraint system.  Granted, the DGI stuff seems very constrained, too.  But within the constraints, it seems unsupervised.  Perhaps the difference lies only in whatever dynamism/reactivity the constraints have?

On 10/26/18 11:07 AM, Marcus Daniels wrote:
> Glen writes:
> 
>    And FWIW, your "scaffolding" sounds a lot like "bootstrapping" <https://en.wikipedia.org/wiki/Bootstrapping_(compilers)> to me.
> 
> It sounds to me like the modularity and re-use of neural nets that Roger directed us to.  


-- 
☣ uǝlƃ



More information about the Friam mailing list