Artem Kaznatcheev Profile picture
Use cancer to know biology. View learning, evolution & phil sci through algorithmic lens. @JSMF Fellow at @Penn, College Lecturer in CS at Oxford. On job market
28 Nov 20
Elimination of mandatory retirement at 65 is a factor leading to bad academic job market: sciencemag.org/careers/2018/1…

Solution: semi-retirement.

"[Having waited to semi-retire until] 74, I in essence removed 9 years from someone else’s career. I should have stepped aside sooner."
I'd be interested to see a graph of average age of retirement for tenured academics on the same plot as the average age of starting a tenured job over the last decades.

I should probably find Larson's paper and see if that sort of data is there.
The paper is here: pubsonline.informs.org/doi/abs/10.128…

Unfortunately, not very big on data. So it doesn't have the figure I want, but the closest is their simulation of number of years as faculty in the lower figure. Image
Read 4 tweets
18 May 20
Late last week, Dave Cohen, Pete Jeavons & I updated our paper on what we can learn about easy vs hard fitness landscapes by studying structure of gene-interaction networks (VCSP instances) that represent them: arxiv.org/abs/1907.01218

Now with many more pictures; prettier, too! ImageImageImageImage
I've already made a thread about the previous version of this work that was published in CP2019:

But if there is interest then I could provide some tweets on the new results and changes in this version.
The final version of our paper looking at the kinds of structure of gene-interaction networks that guarantee easy fitness landscapes is finally out: arxiv.org/abs/1907.01218…

It should appear in JAIR soon, and I think I'll do an updated thread on some of the new results then.
Read 4 tweets
5 Mar 19
Local peaks can't always be found quickly! Hard landscapes are subject to ultimate constraint on evolution: computation. Can hide winding paths.

genetics.org/content/early/…

My path to this paper has been very long: ~7 years in the making. It finally found its peak in @GeneticsGSA.
I'm glad to see this paper is popular with twitter. Thanks!

Since I don't have a SoundCloud to plug, I thought I'd make a tweetstorm summarizing the main results of this paper.

There is a lot to go through, so I apologize for the length of the thread. But I hope you enjoy!
Evolutionary constrains keep populations away from local optima (peaks) in fitness landscapes.

I introduce a division of constraints into 2 types: proximal & ultimate.

This is related to computer science distinction between algorithms & problems [1/n]: egtheory.wordpress.com/2018/07/24/evo…
Read 36 tweets