## September 30, 2012

### Books to Read While the Algae Grow in Your Fur, September 2012

Attention conservation notice: I have no taste.

This edition delayed by my laziness and disorganization going to Allerton.

David Mitchell, The Thousand Autumns of Jacob De Zoet
Mind candy (gourmet). The first part is the beginning of a nicely-written historical novel. Then it gets really remarkably Gothic, with no warning whatsoever. (Spoilers: gur urebvar vf orgenlrq vagb n zlfgrevbhf zbhagnva noorl shyy bs qrsbezrq jbzra, eha ol erartnqr zbaxf cenpgvpvat oynpx zntvp naq trarenyyl pbzvat nf pybfr gb Fngnavfz nf cbffvoyr sbe Ohqquvfgf, jurer fur vf zranprq jvgu sngrf jbefr guna qrngu; gb fnl abguvat bs frperg zrffntrf, fhogyr cbvfbaf, rynobengr orgenlnyf, naq gur ercrngrq vzcyvpngvba gung gur znfgrezvaq bs ivyynval pna npghnyyl jbex zntvp.) I am not sure how seriously to take these parts, whether they are supposed to be as reliably narrated as the beginning or instead are some sort of fantasy on the part of one of the characters (presumably, Jacob). And then the mode shifts yet again, to a sort of parody of novels of the British navy during the Napoleonic wars a la Patrick O'Brian, and then back to the Gothic.
Over-all, it was OK, but I'd have enjoyed it more if it had felt less like an exercise in confusing the reader about what sort of story was being told, and which set of expectations should guide the reading. If there was a point being made about genre expectations, it eluded me.
(Many of the specific ways in which Mitchell's British officers are racist, like calling themselves "Caucasians" and making a big deal of wanting to bring Japan "into the 19th century", seem anachronistic for 1801; so for that matter does Dr. Martinus's discourse on how science will transform humanity --- the closest nearly-period analog I can think of is the strange vision of biotechnology in D'Alembert's Dream, but really it sounds like the late-19th/early-20th century trope of autonomous technics. On the other hand, there was an American trading ship in Nagasaki in 1801, so maybe Mitchell's other seeming anachronisms have historical warrant.)
Marie Brennan, A Star Shall Fall
Mind candy. "Look, it's Halley's Comet!" "The same comet on which you, the Fairy Queen of England, imprisoned the dragon which caused the Great Fire of London, not realizing that the comet would bring back a very angry dragon?" "It seemed like a good idea at the time..." (Not actual quotations.) Fun; I'll look for more by Brennan.
G. E. R. Lloyd, Cognitive Variations: Reflections on the Unity and Diversty of the Human Mind
An examination of evidence presented for and against variations in basic cognitive processes across the human species, being with color perception and spatial orientation, moving through logic, ideas about self and causation, to really abstract things like notions of health and disease. Each topical chapter is largely self-contained: it begins with a over-all discussion of the problem, looks at claims for universality (largely from cognitive and evolutionary psychologists), then at claims of culturally-specific diversity (largely from anthropologists), followed by an examination of evidence from ancient Greece and ancient China (Lloyd's specialties as a historian), ending with a summing-up pointing out that each of these topics is very complicated and shows a combination of unity in some aspects, diversity across cultures in others, and much more diversity within each culture than the anthropologists have usually been willing to allow. (This last point, of internal diversity, is where Lloyd's command of the Greek and Chinese texts is put to really good use.)
The book is short, clear, sensible, and (usually) generous to all the authors discussed. I will however be churlish, and say that I do not thinking that Lloyd is quite fair to the proponents of universality. The ethnographic evidence, and Lloyd's ancient texts, refer to how people talk about color or causation, consciously, often introspectively. But often, I think, the psychologists are interested in more-or-less automatic processing which is unconscious or only barely conscious, hard to put into words, and hard to introspect --- the way we're not conscious of our visual blind-spot, or the grammatical and phonological rules of our native language. (This is part of the point of experiments on pre-verbal infants.) It's hardly inconceivable that while conscious, verbally articulated notions of causation are remarkably various, every human being (or corvid) acts on a similar concept of causality. That said, sometimes the claims of universality are for aspects of cognition which are verbally articulated and accessible to introspection, like color names, and there Lloyd has many shrewd points to make.
Naomi Novik, Victory of Eagles
Mind candy. In which Novik extricates her protagonists from the predicament she left them in at the end of the last book, ultimately dumping them in a situation which is perhaps as bad. I listened to the audiobook this time, and the fact that it had the same narrator as the Aubrey-Maturin series was... interesting.
Elizabeth Bear and Sarah Monette, "Mongoose", "Boojum", and "The Wreck of the Charles Dexter Ward"
Lovecraftian space opera, by way of Lewis Carroll.
Catherynne M. Valente, Silently and Very Fast
If you care for science fiction at all, drop what you're doing and read this. It's one of the best stories about AI — and personal identity, and family continuity, and embodiment and incarnation, and both humanity and post-humanity — I've ever read. The way it re-works mythologies is both brilliant and essential to the story (but it's by Valente, so you knew those things).
This was up for a Hugo award this year, but lost. There is no justice.
Daphne du Maurier, Rebecca
I fail to see what the excitement is about. (This makes more sense to me than other attempts, and it's pretty weak.) There is a great deal of precisely-rendered social embarrassment, even humiliation, but that usually doesn't inspire enthusiasm for a novel. And psychological suspense this was not, since it was perfectly clear to me from the beginning what was going on, and the narrator's failure to get it just made her irritating. Or rather, it was one of many things which made her irritating.
Spoilers: Gur Ovt Frpergf ner gung qrfcvgr (be creuncf orpnhfr bs) ure tbbq gnfgr va vagrevbe qrpbengvba, Erorppn jnf n $x$-gvzvat ovgpu, $x \geq 3$, naq gung Znkvz fubg ure fvapr (guvf vf abg dhvgr ubj gur aneengbe chgf vg) ur jnf gbb zhpu bs n pbjneq gb snpr n qvibepr; ohg nyfb fur qvqa'g zvaq (!) orpnhfr fur'q whfg yrnearq fur unq pnapre. Rkprcg sbe gur pnapre cneg, guvf jnf boivbhf gb zr sebz orsber gur aneengbe naq Znkvz unq ubbxrq hc.
— Did du Maurier influence Shirley Jackson? There is something about the narrator's awkwardness, and her tendency to dwell in spur-of-the-moment fantasies, which reminds me of Eleanor from The Haunting of Hill House.
Mehryar Mohri, Afshin Rostamizadeh and Ameet Talwalkar, Foundations of Machine Learning
Shorter me: A fairly gentle introduction to the mathematical theory of machine learning, focusing on proving performance guarantees, using up-to-date algorithms and problem situations, and showing the tools and techniques used to deliver such proofs more generally. I can't think of any other book now on the market which I'd rather use to teach a first course in learning theory*.
Full review: Machine Learning as Normal Science.
*: The book Foundations most resembles is actually Kearns and Vazirani — which is even more computational, and, being from 1994, leaves two decades of active work out of the picture. While I like Vidyasagar's Learning and Generalization, it's more oriented towards probability and control, and omits contemporary topics like support vector machines, boosting, ranking, and on-line learning. It also costs about three times (!) as much as Foundations. The Elements of Statistical Learning, is great and has more comprehensive coverage (except for ranking), but it's much more methodological and much less proof-oriented. I know that several people are working on textbooks on learning theory from a statistical viewpoint; I urge them to publish soon.
Richard Powers, Gain
The great novel of American corporate capitalism, and how it grew from, and shaped, the lives and hopes of the American people. It is also, and not at all by coincidence or just to jerk tears (though it does), about dying from cancer. Powers's prose is dense and it takes me some effort to get into it, but once I do it is so enthralling that it is very, very worth it.
(First read back in 1998, re-visited now as part of purging my library. This one is a keeper.)

Posted at September 30, 2012 23:59 | permanent link