Re: some ideas

[Joan Pons Semelis <joan@xxxxxxxxxxxxxxxx>]

From: Heikki Levanto <heikki@xxxxxxxxxxxxxxxxx>

> Peter Hallberg (hallberg@xxxxxxxxxxxxxxxxx) wrote in lsd.compgo:

> : In training neural networks it is essential that most of the game space is
> : represented in neurons. 

If this is true then we need a lot of neurons ! 
Even after taking away illegal positions, taking into account simmetries,
and reducing the estimate game space as much as possible, it's still
far too big. Guess we need more neurons than there are electrons in the
universe :-(

Joan