I wish I cold say more than that Christian Schafmeister's work here incredibly interesting... But there's some much in this talk, that I'm kind of at a loss for words.
Have you tried binding two attach points together via a joint molecule, to limit the degrees of freedom for the molecule? It'd be nice to limit the molecule to fewer degrees of freedom, thus reducing the search. As a programmer, limiting it to one axis of rotation would really make it something I can grasp, 2D. Then it would be closely resemble a game like "World of Goo." :)
+Doodle Books - it's the essence of our approach to limit the degrees of freedom of the molecules by creating fused ring, ladder molecules and then connecting the ladders together to create larger structures. The difficulty of the synthesis of the molecules goes up fast the more links you put between things and more than two or three become prohibitive. Synthetic chemistry is a limited and frustrating way of building things but it's all we've got until we can create molecules that create other molecules for us.
+TMH - I'm not sure how machines can learn to design molecules that do things. It's like asking why don't we use machine learning to construct robots or write accounting software or video games. These are complex, creative endeavors - can machine learning do these things?
I wish I cold say more than that Christian Schafmeister's work here incredibly interesting... But there's some much in this talk, that I'm kind of at a loss for words.
Have you tried binding two attach points together via a joint molecule, to limit the degrees of freedom for the molecule? It'd be nice to limit the molecule to fewer degrees of freedom, thus reducing the search.
As a programmer, limiting it to one axis of rotation would really make it something I can grasp, 2D. Then it would be closely resemble a game like "World of Goo." :)
+Doodle Books - it's the essence of our approach to limit the degrees of freedom of the molecules by creating fused ring, ladder molecules and then connecting the ladders together to create larger structures. The difficulty of the synthesis of the molecules goes up fast the more links you put between things and more than two or three become prohibitive. Synthetic chemistry is a limited and frustrating way of building things but it's all we've got until we can create molecules that create other molecules for us.
I wonder why he has chosen to not use machine learning. I'm certain he has good reason but I'm just curious
+TMH - I'm not sure how machines can learn to design molecules that do things. It's like asking why don't we use machine learning to construct robots or write accounting software or video games. These are complex, creative endeavors - can machine learning do these things?