These are my recent Pin​board​.in links:

• Cin­derella Doc­u­men­ta­tion : The­o­ret­i­cal Background

  “At first sight it is not clear whether this require­ment is sat­is­fi­able in gen­eral. Turn on your favorite sys­tem for doing inter­ac­tive geom­e­try or para­met­ric CAD and make the fol­low­ing exper­i­ment: Draw a hor­i­zon­tal line and con­struct two cir­cles of equal radius whose cen­ters are con­strained to slide along the line. Move the cir­cles to a posi­tion in which they inter­sect, and con­struct the upper point of inter­sec­tion of the two cir­cles. Now move one cir­cle so that its cen­ter passes through the cen­ter of the other cir­cle. Most prob­a­bly you will see that the point of inter­sec­tion sud­denly jumps from the upper inter­sec­tion to the lower one. This is what has hap­pened in all the sys­tems we have tried so far. Such behav­ior runs counter to our require­ment of con­ti­nu­ity: You make a small move, and a depen­dent point sud­denly jumps.”
  
  sim­u­la­tion geom­e­try algo­rithms rep­re­sen­ta­tion nudge-​​targets

• Reli­gion and the City | New​geog​ra​phy​.com

  “He talks about items rang­ing from mul­ti­cul­tural sen­si­tiv­i­ties to tak­ing the arts seri­ous to “being famous for help­ing the poor.” The lat­ter was an item that jumped out at me because, as I’ve noted before, too many urban­ist argu­ments are basi­cally argu­ments for what I call “Star­bucks urban­ism.” If called on this, peo­ple will say, “But of course tran­sit will ben­e­fit the poor too.” But that’s not how it’s sold. Urban­ists ought to be famous for the way they design, imple­ment, and talk about their poli­cies as instru­ments for help­ing the poor and facil­i­tat­ing upward eco­nomic and social mobil­ity. There’s a lot of other good stuff in the video that’s rel­e­vant to urban­ism. For those who pre­fer read­ing, Keller also wrote a paper called “Our New Global Cul­ture: Min­istry in Cities, which says of itself: “This paper sur­veys the rise of global cities, the cul­ture and dom­i­nant world­views within these cities, and a frame­work for min­is­ter­ing in them.”?
  
  city-​​planning orga­ni­za­tion mar­ket­ing Workan­tile Coscience out­reach diver­sity man­age­ment

• Study Hacks » Blog Archive » What You Know Mat­ters More Than What You Do

  “Accord­ing to my col­leagues, this star researcher tends to begin with tech­niques, not prob­lems. He first mas­ters a tech­nique that seems promis­ing (and when I say “mas­ter,” I mean it — he really goes deep in build­ing his under­stand­ing). He then uses this new tech­nique to seek out prob­lems that were once hard but now yield eas­ily. He’s rest­less in this quest, often mas­ter­ing sev­eral new tech­niques each year.”
  
  heuris­tics work­life inno­va­tion pro­duc­tiv­ity problem-​​seeing problem-​​solving

• Jour­natic worker takes ‘This Amer­i­can Life’ inside out­sourced jour­nal­ism | Poynter.

  “If you’ve never heard of Jour­natic, that’s kind of the idea. The com­pany, which was founded in 2006, has a web­site that doesn’t appear on at least the first five pages of Google search results. Job open­ings, often posted on Craigslist or Jour​nal​is​mJobs​.com, once men­tioned the company’s name, but no longer. Jour­natic cur­rently works with “dozens” of media com­pa­nies, Tim­pone said, though he declined to name them. He’s spo­ken before of the real estate sec­tion Jour­natic pro­duces for the San Fran­cisco Chron­i­cle. He said more are sign­ing up all the time.”
  
  disintermediation-​​in-​​action jour­nal­ism work­life out­sourc­ing exposé

• A Step-​​by-​​Step Guide to Tribal Lead­er­ship: Part 1: The Five Stages of Tribal Cul­ture « emer­gent by design

  “Tribal Lead­ers are the peo­ple who focus their efforts on upgrad­ing the tribal cul­ture. (upgrad­ing the words we use to describe our real­ity and the behav­iors we prac­tice that shape the direc­tion of our lives) They set the stan­dard of per­for­mance in their indus­tries, from pro­duc­tiv­ity and prof­itabil­ity to employee reten­tion, and attract tal­ent. Most of all, they help bring groups to unity by rec­og­niz­ing their ‘trib­al­ness’ – get­ting peo­ple to talk about the things they really care about, com­ing together around these com­mon causes, and form­ing mis­sions to make some­thing great hap­pen, and to live in great­ness. The goal of Tribal Lead­er­ship is to learn how to get peo­ple ‘unstuck’ – from unhelp­ful lan­guage and behav­iors, so we can level up and tran­si­tion into higher-​​performance, less stress­ful, and more fun states of Being.”
  
  i-​​hate-​​the-​​word-​​tribes col­lab­o­ra­tion lead­er­ship cultural-​​dynamics advice

• [1206.6532] Esti­mat­ing Nui­sance Para­me­ters in Inverse Problems

  “Many inverse prob­lems include nui­sance para­me­ters which, while not of direct inter­est, are required to recover pri­mary para­me­ters. Struc­ture present in these prob­lems allows effi­cient opti­miza­tion strate­gies — a well known exam­ple is vari­able pro­jec­tion, where non­lin­ear least squares prob­lems which are lin­ear in some para­me­ters can be very effi­ciently opti­mized. In this paper, we extend the idea of pro­ject­ing out a sub­set over the vari­ables to a broad class of max­i­mum like­li­hood (ML) and max­i­mum a pos­te­ri­ori like­li­hood (MAP) prob­lems with nui­sance para­me­ters, such as vari­ance or degrees of free­dom. As a result, we are able to incor­po­rate nui­sance para­me­ter esti­ma­tion into large-​​scale con­strained and uncon­strained inverse prob­lem for­mu­la­tions. We apply the approach to a vari­ety of prob­lems, includ­ing esti­ma­tion of unknown vari­ance para­me­ters in the Gauss­ian model, degree of free­dom (d.o.f.) para­me­ter esti­ma­tion in the con­text of robust inverse prob­lems, auto­matic cal­i­bra­tion, and opti­mal exper­i­men­tal design. Using numer­i­cal exam­ples, we demon­strate improve­ment in recov­ery of pri­mary para­me­ters for sev­eral large– scale inverse prob­lems. The pro­posed approach is com­pat­i­ble with a wide vari­ety of algo­rithms and for­mu­la­tions, and its imple­men­ta­tion requires only minor mod­i­fi­ca­tions to exist­ing algorithms.”
  
  reinventing-​​the-​​wheel feature-​​extraction opti­miza­tion modeling-​​is-​​not-​​mathematics nudge-​​targets

• [1206.4608] A Hybrid Algo­rithm for Con­vex Semi­def­i­nite Optimization

  “We present a hybrid algo­rithm for opti­miz­ing a con­vex, smooth func­tion over the cone of pos­i­tive semi­def­i­nite matri­ces. Our algo­rithm con­verges to the global opti­mal solu­tion and can be used to solve gen­eral large-​​scale semi­def­i­nite pro­grams and hence can be read­ily applied to a vari­ety of machine learn­ing prob­lems. We show exper­i­men­tal results on three machine learn­ing prob­lems (matrix com­ple­tion, met­ric learn­ing, and sparse PCA) . Our approach out­per­forms state-​​of-​​the-​​art algorithms.”
  
  algo­rithms opti­miza­tion computational-​​complexity spe­cial­iza­tion nudge-​​targets

• [1206.6690] Gen­er­a­tion and Prop­er­ties of Snarks

  “For many of the unsolved prob­lems con­cern­ing cycles and match­ings in graphs it is known that it is suf­fi­cient to prove them for emph{snarks}, the class of non­triv­ial 3-​​regular graphs which can­not be 3-​​edge coloured. In the first part of this paper we present a new algo­rithm for gen­er­at­ing all non-​​isomorphic snarks of a given order. Our imple­men­ta­tion of the new algo­rithm is 14 times faster than pre­vi­ous pro­grams for gen­er­at­ing snarks, and 29 times faster for gen­er­at­ing weak snarks. Using this pro­gram we have gen­er­ated all non-​​isomorphic snarks on $nleq 36$ ver­tices. Pre­vi­ously lists up to $n=28$ ver­tices have been pub­lished. In the sec­ond part of the paper we ana­lyze the sets of gen­er­ated snarks with respect to a num­ber of prop­er­ties and con­jec­tures. We find that some of the strongest ver­sions of the cycle dou­ble cover con­jec­ture hold for all snarks of these orders, as does Jaeger’s Petersen colour­ing con­jec­ture, which in turn implies that Fulkerson’s con­jec­ture has no small coun­terex­am­ples. In con­trast to these pos­i­tive results we also find coun­terex­am­ples to eight pre­vi­ously pub­lished con­jec­tures con­cern­ing cycle cov­er­ings and the gen­eral cycle struc­ture of cubic graphs.”
  
  graph-​​theory com­bi­na­torics algo­rithms nudge-​​targets

• [1206.6238] Entrain­abil­ity enhance­ment by period mis­match in biloop genetic oscillators

  “Effects of the period mis­match on entrain­ment prop­er­ties in two cou­pled genetic oscil­la­tors are stud­ied. The entrain­ment is cal­cu­lated with a phase reduc­tion approach and a Flo­quet mul­ti­plier analy­sis, and their depen­den­cies on cou­pling strength and the period ratio are inves­ti­gated in two genetic oscil­la­tor mod­els (smooth and relax­ation oscil­la­tors). We find that the exis­tence of the period mis­match induces an enhance­ment of entrain­ment in both smooth and relax­ation oscil­la­tors. By cal­cu­lat­ing Flo­quet mul­ti­pli­ers, we show that the enhance­ment mech­a­nism is based on the cou­pled oscil­la­tors which are in the vicin­ity of bifur­ca­tion on limit cycle.”
  
  biological-​​engineering emergent-​​design reaction-​​networks oscil­la­tors control-​​theory

• [1206.4672] Effi­cient Active Algo­rithms for Hier­ar­chi­cal Clustering

  “Advances in sens­ing tech­nolo­gies and the growth of the inter­net have resulted in an explo­sion in the size of mod­ern datasets, while stor­age and pro­cess­ing power con­tinue to lag behind. This moti­vates the need for algo­rithms that are effi­cient, both in terms of the num­ber of mea­sure­ments needed and run­ning time. To com­bat the chal­lenges asso­ci­ated with large datasets, we pro­pose a gen­eral frame­work for active hier­ar­chi­cal clus­ter­ing that repeat­edly runs an off-​​the-​​shelf clus­ter­ing algo­rithm on small sub­sets of the data and comes with guar­an­tees on per­for­mance, mea­sure­ment com­plex­ity and run­time com­plex­ity. We instan­ti­ate this frame­work with a sim­ple spec­tral clus­ter­ing algo­rithm and pro­vide con­crete results on its per­for­mance, show­ing that, under some assump­tions, this algo­rithm recov­ers all clus­ters of size ?(log n) using O(n log^2 n) sim­i­lar­i­ties and runs in O(n log^3 n) time for a dataset of n objects. Through exten­sive exper­i­men­ta­tion we also demon­strate that this frame­work is prac­ti­cally alluring.”
  
  clus­ter­ing algo­rithms nudge-​​targets practically-​​alluring

• Most Bla­tant Pro-​​ACTA Cam­paign So Far Is A Copy­right Monop­oly Vio­la­tion — Falkvinge on Infopolicy

  “This episode shows clearer than ever that the copy­right and patent monop­o­lies are not intended to be pro­tec­tive of inno­va­tion or pro­tec­tive of the econ­omy. They’re obvi­ously too com­plex even for their strongest sup­port­ers and lob­by­ists to under­stand and adhere to. Rather, they are intended as legal clubs to be used by the now-​​rich incum­bents against resource-​​strapped upstarts. The copy­right and patent monop­o­lies are only pro­tec­tive of the past, pro­tec­tive against the present and future of inno­va­tion, cre­ativ­ity, and economy.”
  
  copy­right intellectual-​​property cor­po­ratism public-​​policy

• [1112.5218] Pat­terns of neu­tral diver­sity under gen­eral mod­els of selec­tive sweeps

  “Two major sources of sto­chas­tic­ity in the dynam­ics of neu­tral alle­les result from resam­pling of finite pop­u­la­tions (genetic drift) and the ran­dom genetic back­ground of nearby selected alle­les on which the neu­tral alle­les are found (linked selec­tion). There is now good evi­dence that linked selec­tion plays an impor­tant role in shap­ing poly­mor­phism lev­els in a num­ber of species. One of the best inves­ti­gated mod­els of linked selec­tion is the recur­rent full sweep model, in which newly arisen selected alle­les fix rapidly. How­ever, the bulk of selected alle­les that sweep into the pop­u­la­tion may not be des­tined for rapid fix­a­tion. Here we develop a gen­eral model of recur­rent selec­tive sweeps in a coa­les­cent frame­work, one that gen­er­al­izes the recur­rent full sweep model to the case where selected alle­les do not sweep to fix­a­tion. We show that in a large pop­u­la­tion, only the ini­tial rapid increase of a selected allele affects the geneal­ogy at par­tially linked sites, which under fairly gen­eral assump­tions are unaf­fected by the sub­se­quent fate of the selected allele. We also apply the the­ory to a sim­ple model to inves­ti­gate the impact of recur­rent par­tial sweeps on lev­els of neu­tral diver­sity, and find that for a given reduc­tion in diver­sity, the impact of recur­rent par­tial sweeps on the fre­quency spec­trum at neu­tral sites is deter­mined pri­mar­ily by the fre­quen­cies achieved by the selected alle­les. Con­se­quently, recur­rent sweeps of selected alle­les to low fre­quen­cies can have a pro­found effect on lev­els of diver­sity but can leave the fre­quency spec­trum rel­a­tively unper­turbed. In fact, the lim­it­ing coa­les­cent model under a high rate of sweeps to low fre­quency is iden­ti­cal to the stan­dard neu­tral model. The gen­eral model of selec­tive sweeps we describe goes some way towards pro­vid­ing a more flex­i­ble frame­work to describe genomic pat­terns of diver­sity than is cur­rently available.”
  
  neutral-​​networks evolutionary-​​dynamics fitness-​​landscapes diver­sity theoretical-​​biology

• [1206.3520] Recov­er­ing the tree-​​like trend of evo­lu­tion despite exten­sive lat­eral genetic trans­fer: A prob­a­bilis­tic analysis

  “In the pres­ence of high­ways, deal­ing with more gen­eral net­work set­tings would be desir­able. Also our def­i­n­i­tion of high­ways as con­nect­ing two edges is some­what restric­tive. In gen­eral, one is also inter­ested in pref­er­en­tial genetic trans­fers between clades.”
  
  algo­rithms lateral-​​gene-​​transfer cladis­tics phy­lo­ge­net­ics inverse-​​problems ontol­ogy modeling-​​is-​​not-​​mathematics nudge-​​targets

• [1206.3279] The Phy­lo­ge­netic Indian Buf­fet Process: A Non-​​Exchangeable Non­para­met­ric Prior for Latent Features

  “Non­para­met­ric Bayesian mod­els are often based on the assump­tion that the objects being mod­eled are exchange­able. While appro­pri­ate in some appli­ca­tions (e.g., bag-​​of-​​words mod­els for doc­u­ments), exchange­abil­ity is some­times assumed sim­ply for com­pu­ta­tional rea­sons; non-​​exchangeable mod­els might be a bet­ter choice for appli­ca­tions based on sub­ject mat­ter. Draw­ing on ideas from graph­i­cal mod­els and phy­lo­ge­net­ics, we describe a non-​​exchangeable prior for a class of non­para­met­ric latent fea­ture mod­els that is nearly as effi­cient com­pu­ta­tion­ally as its exchange­able coun­ter­part. Our model is applic­a­ble to the gen­eral set­ting in which the depen­den­cies between objects can be expressed using a tree, where edge lengths indi­cate the strength of rela­tion­ships. We demon­strate an appli­ca­tion to mod­el­ing prob­a­bilis­tic choice.”
  
  sta­tis­tics algo­rithms ontol­ogy col­li­ga­tion feature-​​extraction philosophy-​​of-​​science nudge-​​targets

• “The Eurozone’s Strat­egy is a Dis­as­ter” « naked capitalism

  “Why should Ger­man and other tax­pay­ers, mostly from the north, pay for the oth­ers, mostly from the south? Because their gov­ern­ments are respon­si­ble for the dis­as­trous sit­u­a­tion we are in.”
  
  financial-​​crisis public-​​policy eco­nom­ics cultural-​​dynamics fair-​​weather-​​bosses

• [1206.6504] An Abstract Approach to Strat­i­fi­ca­tion in Lin­ear Logic

  “We study the notion of strat­i­fi­ca­tion, as used in sub­sys­tems of lin­ear logic with low com­plex­ity bounds on the cut-​​elimination pro­ce­dure (the so-​​called light log­ics), from an abstract point of view, intro­duc­ing a log­i­cal sys­tem in which strat­i­fi­ca­tion is han­dled by a sep­a­rate modal­ity. This modal­ity, which is a gen­er­al­iza­tion of the para­graph modal­ity of Girard’s light lin­ear logic, arises from a gen­eral cat­e­gor­i­cal con­struc­tion applic­a­ble to all mod­els of lin­ear logic. We thus learn that strat­i­fi­ca­tion may be for­mu­lated inde­pen­dently of expo­nen­tial modal­i­ties; when it is forced to be con­nected to expo­nen­tial modal­i­ties, it yields inter­est­ing com­plex­ity prop­er­ties. In par­tic­u­lar, from our analy­sis stem three alter­na­tive refor­mu­la­tions of Bail­lot and Mazza’s lin­ear logic by lev­els: one geo­met­ric, one inter­ac­tive, and one semantic.”
  
  linear-​​logic logic-​​programming for­mal­iza­tion nudge-​​targets rep­re­sen­ta­tion

• [1205.0802] Win-​​stay-​​lose-​​learn pro­motes coop­er­a­tion in the spa­tial prisoner’s dilemma game

  “Hold­ing on to one’s strat­egy is nat­ural and com­mon if the later war­rants suc­cess and sat­is­fac­tion. This goes against wide­spread sim­u­la­tion prac­tices of evo­lu­tion­ary games, where play­ers fre­quently con­sider chang­ing their strat­egy even though their pay­offs may be mar­gin­ally dif­fer­ent than those of the other play­ers. Inspired by this obser­va­tion, we intro­duce an aspiration-​​based win-​​stay-​​lose-​​learn strat­egy updat­ing rule into the spa­tial prisoner’s dilemma game. The rule is sim­ple and intu­itive, fore­see­ing strat­egy changes only by dis­sat­is­fied play­ers, who then attempt to adopt the strat­egy of one of their near­est neigh­bors, while the strate­gies of sat­is­fied play­ers are not sub­ject to change. We find that the pro­posed win-​​stay-​​lose-​​learn rule pro­motes the evo­lu­tion of coop­er­a­tion, and it does so very robustly and inde­pen­dently of the ini­tial con­di­tions. In fact, we show that even a minute ini­tial frac­tion of coop­er­a­tors may be suf­fi­cient to even­tu­ally secure a highly coop­er­a­tive final state. In addi­tion to exten­sive sim­u­la­tion results that sup­port our con­clu­sions, we also present results obtained by means of the pair approx­i­ma­tion of the stud­ied game. Our find­ings con­tinue the suc­cess story of related win-​​stay strat­egy updat­ing rules, and by doing so reveal new ways of resolv­ing the prisoner’s dilemma.”
  
  game-​​theory agent-​​based com­plex­ol­ogy

• The Rude Pundit

  “And there’s every­thing you need to know about the Repub­li­can Party. “Shit hap­pened, but so what? Peo­ple were vic­tim­ized, but why should we care? That was nearly forty years ago.” The demen­tia in refus­ing to look back­ward, refus­ing to make up for the mis­takes of the past, whether it’s the Bush tax cuts or the lies that got us into war or the lies that got us into this finan­cial cri­sis, makes us damned to repeat. ”
  
  sum­mary pol­i­tics Repub­li­cans