May 31, 2019

Books to Read While the Algae Grow in Your Fur, May 2019

Attention conservation notice: I have no taste, and no qualifications to opine on international political economy and the global financial crises of 2008--, social theory, the history of the Roman Empire, or even the history of statistics. Also, I seem to have done a lot of re-reading this month.

Linda Nagata, Vast
Science fiction, rising above the level of mind candy. I read this twenty years ago, and re-read it this spring because the sequel has, at last, come out. It's just as good as I recalled at conveying the sense of wonder and desolation, the scale and age and strangeness of the universe. What my memory hadn't retained was how Nagata combines both an unsparing, utterly unsentimental narration with many genuinely affecting --- I hesitate to say "human" --- experiences. (That last might be because I am now middle aged, rather than in the prolonged adolescence of graduate school, and have more basis on which to relate to those moments.) Also, the moments of horror are a lot more effectively horrifying than I recalled. But if I try to go further in this line I will merely repeat what Henry said, only less eloquently. I will just say that this deserves to be a classic of science fiction, and that if you like well-written hard (*) SF, you owe it to yourself to read this. §
*: As for the hardness of the SF, Nagata doesn't play any tricks with special relativity; the closest she comes is a slower-than-light drive that effectively tilts gravity locally. (This would of course have interesting general-relativistic implications.) Drexler-style molecular nanotechnology and mind-uploading are big parts of the book, and there are of course disputes over whether those are scientifically viable, but I am inclined to accept them as "very difficult but not flat-out impossible". This makes the book harder SF than pretty much all of Niven and a good deal of Egan.
While on this subject: Henry describes Vast's "philosopher cells" as "cellular automata turned lethal". I am prepared to go further --- I'm pretty sure Nagata was directly inspired by popular accounts (perhaps in Levy's Artificial Life, or Poundstone's [also-ought-to-be-a-classic] Recursive Universe) of cellular automata as distributed, self-reproducing systems. Indeed, some aspects of her depictions of Lot's attempts to sway the philosopher cells make me think that she played around with CA simulators. On the other hand, I think her depictions of evolutionary design derive from Drexler's Engines of Creation, rather than, say, SFI's Evolving Cellular Automata Project or even John Holland.
Robyn Bennis, The Devil's Guide to Managing Difficulty People
Mind candy: contemporary comic fantasy, featuring central Florida, Boston, and Hell. (Obvious jokes are obvious, and made in the book.) §
Judy L. Klein, Statistical Visions in Time: A History of Time Series Analysis, 1662--1938
While it's 20 years old now, I'm not aware of anything which can really replace or even complement it as a history of the development of time series up to the self-conscious introduction of the ideas of stationary stochastic processes and the Wold decomposition. I recommend it unreservedly for anyone seriously interested in time series analysis.
There are two parts. Part I looks at the roots of common time series techniques in non-scientific mathematical practice, especially in business and finance. This includes things like differencing series; using index numbers and looking at relative changes; moving averages; transposing series into "cycle time" and averaging to extract typical seasonal (or longer-term) patterns; etc. In each case, Klein documents a progression from "commercial arithmetic" or "political arithmetic" to, gradually, the use of these techniques for scientific investigation. (There are some demi-Marxist hints here about social structure controlling what people are able to perceive. Klein does not follow up on these hints, which is just as well.)
Part II traces the path through which statistical and probabilistic concepts which were originally developed for looking at cross-sections through a population at one time, such as probability distributions (especially the Gaussian distribution), correlation, and regression eventually became the tools used to relate successive values within a single time series, with no obvious population in sight. Klein succeeds in making this seem strange, which is a real and important accomplishment. I do not think it's necessarily quite as weird as she'd make it out to be, at least when the series is statistically stationary, because then it's essentially licensed by ergodicity. That said, her core complaint is that to get to a stationary process, we use all the tricks from Part I, which lack the sound mathematical basis that (on her telling) the Russians like Markov, Khinchin and Kolmogorov gave to stationary processes, and in doing so we may end up discarding what's most interesting in the data. §
*: I do find it a little strange that her account of the development of stochastic processes says almost nothing about physics (as opposed to meteorology). There's a bit about Maxwell and Boltzmann's ideas of statistical mechanics and how they were influenced by Quetelet, and there's mention of Einstein and Smoluchowski, but no recognition of how they described Brownian motion as a stochastic process, nor any mention of Langevin or the Ehrenfests, and I think Wiener is only mentioned once (in a quotation), the ideas of ergodic theory aren't discussed, etc. One could of course argue about how much these fed in to the development of time series analysis in statistics and economics (Klein's main interest), but I think even there there's a link, via the influence on Khinchin and Kolmogorov (who were certainly very aware of all this). Cf. von Plato's Creating Modern Probability (which I really ought to finish and review one of these years).
Brian Daley, The Doomfollowers of Coramonde
Mind candy: 1970s-vintage secondary-world epic fantasy, which is also a portal fantasy with a US Army armored personnel carrier sorcerously detached from duty in Vietnam. It's enjoyable in its own terms, but three things strike me about it. First, all of the characters have agency --- the inhabitants of the fantasy world don't just react to the Americans who come through the portal, they pursue their own lives and agendas. Second, while the female characters do have agency, wow do we hear an awful lot about their looks. Third, this is incredibly compressed by present standards --- there are multiple conceits in here which would each inspire full trilogies (or longer series) in modern publishing. (I have read such series with enjoyment.)
Picked up because I liked Daley's science fiction as a teenager, and was curious when I ran across this and its sequel (which I'll probably read). §
Tanya Huff, The Better Part of Valor
Mind candy: military SF, space marines, all-commissioned-officers-are-idiots variety. §
James J. O'Donnell, The Ruin of the Roman Empire: A New History
This is intensely focused on the period from Theodoric (presented as a basically ordinary Roman emperor of Italy, Gaul and the Balkans) through Justinian (the villain of the piece, who dooms the Roman world by attempting to re-unify it) to Pope Gregory the Great (presented as the last consul of the city of Rome). As my parentheticals indicate, O'Donnell has some unusual notions, which he does not hesitate to push very strongly. (These include what sound to me like some very odd judgments about what would constitute geographically natural political territories; explicating the implied causal theories here would be... curious.) He doesn't end up convincing me that everything can be blamed on Justinian and his court, but it's still a well-written and provocative look at a fascinating historical epoch. §
Ernest Gellner, Contemporary Thought and Politics
A collection of Gellner's essays from the 1950s, 1960s and early 1970s. I first read it in 1997 as part of my rush through everything he'd written (after having my mind blown by Nations and Nationalism). Lucking across a used copy, I enjoyed the re-read: Gellner's acerbic wit, playful style, acuity, and characteristic themes are all on good display here, as he touches on student revolt, the Prague Spring, the idea of progress, the relationship between the scale of social integration and nationalism, ideas about the philosophy of history and of social science, how the social organization of Berber tribes illuminates international relations, and so on. But the targets of his writing are sufficiently far in the past, by this point, that I can only recommend it for Gellner completists. (I am one.) If you are not already a committed Gellnerian, go read Nations and Nationalism, or Plough, Sword and Book. §
Goran Peskir, From Uniform Laws of Large Numbers to Uniform Ergodic Theorems (University of Aarhus Department of Mathematics Lecture Notes Series, vol. 66; 2000) [PDF reprint via Prof. Peskir]
This is an advanced monograph for readers who are comfortable with measure-theoretic probability, and ideally some exposure to at least one of uniform laws of large numbers (as in Vapnik-Chervonenkis theory or statistical learning theory), empirical process theory, or the ergodic theory of stochastic processes and dynamical systems. So I won't be very expository in this brief notice.
The book begins with reminders about how we prove laws of large numbers, how uniform laws of large numbers work and how we prove them, and ergodic theory, especially the way we can go back and forth between stochastic-process and dynamical-systems descriptions, and the classic ergodic theorems. The book's main chapters then each cover three ways of extending ergodic theorems to uniform ergodicity, i.e., to ergodic theorems which hold uniformly over all functions in some class:
  1. A very abstract condition based on metric entropy for dynamical systems (and consequently for ergodic processes);
  2. A Vapnik-Chervonenkis condition for beta-mixing (absolutely regular) stochastic processes;
  3. A Fourier-spectral condition for weakly stationary processes.
Approach (2) is now fairly common, following Yu (1994), though Peskir can legitimately claim to have been a pioneer. (His references are very good, up to the time of writing.) Approach (1) was not one I'd seen worked out in quite this way, though it's closely related to work by van Handel (2013). Approach (3) was entirely new to me, and very intriguing.
While there is a lot of repetition of definitions, theorem statements, etc., the line-by-line writing is fine, and the proofs are easy to follow. I'd been meaning to read this since 2003 or 2004, and kept putting it off; this was my loss. If you're seriously interested in uniform convergence and/or statistical learning theory for stochastic processes, this is very worthwhile. §
Dorothy Sayers, Whose Body?; Clouds of Witness; The Unpleasantness at the Bellona Club; Unnatural Death; Strong Poison; The Five Red Herrings; Have His Carcase
I was introduced to Sayers's mystery novels as a boy by my mother. Re-reading these first seven in a burst, they're still great. They helped define what detective stories could be, and, almost a century later, what they too rarely are. §
Adam Tooze, Crashed: How a Decade of Financial Crises Changed the World
This is a great history of what lead up to the 2008 financial crisis, and its effects down, basically, to the time of writing. What makes it so good is that Tooze has a mastery of the material on, it seems, every possible scale. Whether he is writing about the mechanics of all kinds of financial markets and institutions, the personalities and interactions of crucial decision makers, broad climates of opinion, geopolitical rivalries or low-frequency economic trends, he is full of judiciously-chosen details, lucidly explained, and is clear on the inter-relations. (He is also accurate about areas where I have independent knowledge.) Reading him is such a pleasure that it sometimes was hard for me to take in his larger message, which is that the world went through an entirely-preventable disaster, and only managed to avoid an even worse catastrophe because of luck, dubiously-legitimate actions by unaccountable bureaucrats, and lingering sentimental attachment to institutions and alliances which have become hollowed out since the end of the Cold War. The fact that America and the US government are still essential to the global economy, and what there is of a global polity, but that our internal politics are so broken we have to hope that the Fed will do what our elected leaders can't or won't, is a running theme. (And of course the Fed doesn't even have the representative agent's welfare in view.) The horrid repercussions of these disasters, and the just-barely-enough measures to counter them, have left us in a position where the next disaster will be even harder to deal with --- and it will come. §

Books to Read While the Algae Grow in Your Fur; Scientifiction and Fantastica; Pleasures of Detection, Portraits of Crime; Writing for Antiquity; Enigmas of Chance; The Dismal Science; Philosophy; Commit a Social Science; The Continuing Crises; Data Over Space and Time

Posted at May 31, 2019 23:59 | permanent link

Three-Toed Sloth