Algorithm Reveals Secrets of Leaf Shape

technologyreview.com

23 points·instantramen··5 comments

Bonus URL, on parameter tuning: http://www.hvass-labs.org/people/magnus/thesis/pedersen08the... (PDF, Tuning And Simplifying Heuristical Optimization (PhD thesis)). Via http://en.wikipedia.org/wiki/Particle_swarm_optimization

Someone may want to "cross-breed" those papers. I found Pedersen's interesting thesis via the picture he kindly contributed to that Wikipedia page (thank you; I used it for harmless, I hope, pedagogic purposes).

Good find. Snippet:

In his model, the growth of leaf lobes is governed by the position of leaf veins. [...] By varying just a handful of parameters, Young can produce a surprisingly rich variety of shapes.

The issue of the shapes not found in nature is also thought-provoking ...

P.S. Technology Review editors: http://begthequestion.info/ , please. Thank you.

I read the article diagonally, and found no samples of what non-natural leaf shapes does the algorithm produce. Sad. But it is interesting that even complicated shapes like maple can be produced with such a simple model.

The editors' "produces one or two shapes not found in nature" may mean Fig.5a and Fig.6c, which are lacking the '[as] seen in' notes given for all others presented.

I'd say the author was going for showing that the model covered many (all ? ferns, etc.?) natural shapes. But both the issues of exists-in-nature-but-not-in-model and vice-versa should be interesting.

Perhaps the sets of "these parameters do not show up in nature" can give the bio-mechanisms researchers some hints.

I think those missing sets can be very interesting. I didn't realise these figures didn't have tags (I scanned the text twice, but missed them :/)

The problem of lack-of-space in journals justifies these kind of cuts, but you can always submit a longer version to ArXiV without any problems or restrictions

Reply on news.ycombinator.com