“The extremal index para­me­ter theta char­ac­ter­izes the degree of local depen­dence in the extremes of a sta­tion­ary time series and has impor­tant appli­ca­tions in a num­ber of areas, such as hydrol­ogy, telecom­mu­ni­ca­tions, finance and envi­ron­men­tal studies.…Further, a pro­ce­dure for the auto­matic selec­tion of its tun­ing para­me­ter is devel­oped and dif­fer­ent types of con­fi­dence inter­vals that prove use­ful in prac­tice pro­posed. The per­for­mance of the esti­ma­tor is exam­ined through sim­u­la­tions, which show its highly com­pet­i­tive behav­ior. Finally, the esti­ma­tor is applied to three real data sets of daily crude oil prices, daily returns of the S&P 500 stock index, and high-​​frequency, intra-​​day traded vol­umes of a stock. These appli­ca­tions demon­strate addi­tional diag­nos­tic fea­tures of sta­tis­ti­cal plots based on the new estimator.”

“I am lis­ten­ing to a pre­sen­ta­tion at the Homer Hoyt meet­ings on the condo melt­down in South Florida. Devel­op­ers planned on build­ing 95,000 units in the city of Miami between 2002 and 2007. In the 2000 cen­sus, the whole city had 163,000 units.”

“Research into prop­er­ties of het­ero­ge­neous arti­fi­cial mate­ri­als, con­sist­ing of arrange­ments of rigid scat­ter­ers embed­ded in a medium with dif­fer­ent elas­tic prop­er­ties, has been intense through­out last two decades. The capa­bil­ity to pre­vent the trans­mis­sion of waves in pre­de­ter­mined bands of fre­quen­cies –called bandgaps– becomes one of the most inter­est­ing prop­er­ties of these sys­tems, and leads to the pos­si­bil­ity of design­ing devices to con­trol wave prop­a­ga­tion. The under­ly­ing phys­i­cal mech­a­nism is destruc­tive Bragg inter­fer­ence. Here we show a tech­nique that enables the cre­ation of a wide bandgap in these mate­ri­als, based on frac­tal geome­tries. We have focused our work in the acoustic case where these mate­ri­als are called Phononic/​Sonic Crys­tals (SC) but, the tech­nique could be applied any types of crys­tals and wave types in ranges of fre­quen­cies where the physics of the process is linear.”

“The opti­miza­tion for­mu­la­tion con­sid­ered in this paper uses a penal­ized neg­a­tive Pois­son log-​​likelihood objec­tive func­tion with non­neg­a­tiv­ity con­straints (since Pois­son inten­si­ties are nat­u­rally non­neg­a­tive). In par­tic­u­lar, the pro­posed approach incor­po­rates key ideas of using sep­a­ra­ble qua­dratic approx­i­ma­tions to the objec­tive func­tion at each iter­a­tion and penal­iza­tion terms related to l1 norms of coef­fi­cient vec­tors, total vari­a­tion semi­norms, and partition-​​based mul­ti­scale esti­ma­tion methods.”

“The obtained results and trends sug­gest a num­ber of fur­ther inves­ti­ga­tions. For instance, it would be inter­est– ing to con­sider other net­work mod­els and mea­sure­ments, as well as to assess the effect of dif­fer­ent types of hard– ware, com­pil­ers and oper­at­ing systems.”

“Ran­dom num­bers play a cru­cial role in sci­ence and indus­try. Many numer­i­cal meth­ods require the use of ran­dom num­bers, in par­tic­u­lar the Monte Carlo method. There­fore it is of para­mount impor­tance to have effi­cient ran­dom num­ber gen­er­a­tors. The dif­fer­ences, advan­tages and dis­ad­van­tages of true and pseudo ran­dom num­ber gen­er­a­tors are dis­cussed with an empha­sis on the intrin­sic details of mod­ern and fast pseudo ran­dom num­ber gen­er­a­tors. Fur­ther­more, stan­dard tests to ver­ify the qual­ity of the ran­dom num­bers pro­duced by a given gen­er­a­tor are out­lined. Finally, stan­dard sci­en­tific libraries with built-​​in gen­er­a­tors are pre­sented, as well as dif­fer­ent approaches to gen­er­ate nonuni­form ran­dom num­bers. Poten­tial prob­lems that one might encounter when using large par­al­lel machines are discussed.”

“In many net­works, ver­tices have hid­den attrib­utes, or types, that are cor­re­lated with the net­works topol­ogy. If the topol­ogy is known but these attrib­utes are not, and if learn­ing the attrib­utes is costly, we need a method for choos­ing which ver­tex to query in order to learn as much as pos­si­ble about the attrib­utes of the other ver­tices. We assume the net­work is gen­er­ated by a sto­chas­tic block model, but we make no assump­tions about its assor­ta­tiv­ity or dis­as­sor­ta­tiv­ity. We choose which ver­tex to query using two meth­ods: 1) max­i­miz­ing the mutual infor­ma­tion between its attrib­utes and those of the oth­ers (a well-​​known approach in active learn­ing) and 2) max­i­miz­ing the aver­age agree­ment between two inde­pen­dent sam­ples of the con­di­tional Gibbs dis­tri­b­u­tion. Exper­i­men­tal results show that both these meth­ods do much bet­ter than sim­ple heuris­tics. They also con­sis­tently iden­tify cer­tain ver­tices as impor­tant by query­ing them early on.”

“We pro­pose a fluc­tu­a­tion analy­sis to quan­tify spa­tial cor­re­la­tions in com­plex net­works. The approach con­sid­ers the sequences of degrees along short­est paths in the net­works and quan­ti­fies the fluc­tu­a­tions in anal­ogy to time series. In this work, the Barabasi-​​Albert (BA) model, the Cay­ley tree at the per­co­la­tion tran­si­tion, a frac­tal net­work model, and exam­ples of real-​​world net­works are stud­ied. While the fluc­tu­a­tion func­tions for the BA model show expo­nen­tial decay, in the case of the Cay­ley tree and the frac­tal net­work model the fluc­tu­a­tion func­tions dis­play a power-​​law behav­ior. The frac­tal net­work model com­prises long-​​range anti-​​correlations. The results sug­gest that the fluc­tu­a­tion expo­nent pro­vides com­ple­men­tary infor­ma­tion to the frac­tal dimension.”

“Wealth is no proof of moral char­ac­ter; nor poverty of the want of it.

On the con­trary, wealth is often the pre­sump­tive evi­dence of dis­hon­esty; and poverty the neg­a­tive evi­dence of inno­cence. If there­fore prop­erty, whether lit­tle or much, be made a cri­te­rion, the means by which that prop­erty has been acquired ought to be made a cri­te­rion also.”

“This means that each time the doc­u­ment is updated, the client will also store the pre­vi­ous ver­sion as an attach­ment to the lat­est ver­sion. At any time, a user can load any of the old versions.”

“Com­mu­nity struc­ture is one of the key prop­er­ties of com­plex net­works and plays a cru­cial role in their topol­ogy and func­tion. While an impres­sive amount of work has been done on the issue of com­mu­nity detec­tion, very lit­tle atten­tion has been so far devoted to the inves­ti­ga­tion of com­mu­ni­ties in real net­works. We present a sys­tem­atic empir­i­cal analy­sis of the sta­tis­ti­cal prop­er­ties of com­mu­ni­ties in large infor­ma­tion, com­mu­ni­ca­tion, tech­no­log­i­cal, bio­log­i­cal, and social net­works. We find that the meso­scopic orga­ni­za­tion of net­works of the same cat­e­gory is remark­ably sim­i­lar. This is reflected in sev­eral char­ac­ter­is­tics of com­mu­nity struc­ture, which can be used as “fin­ger­prints” of spe­cific net­work categories.…”

“… We find that SF net­works gen­er­ate oscil­la­tions much more eas­ily than ER net­works do, and this may explain why SF net­works are more evolv­able than ER net­works are for oscil­la­tory phe­no­types. In spite of their greater evolv­abil­ity, we find that net­works with SFout topolo­gies are also more robust to muta­tions than ER net­works. Fur­ther­more, the SFout topolo­gies are more robust to changes in ini­tial con­di­tions (envi­ron­men­tal robust­ness). For both topolo­gies, we find that once a pop­u­la­tion of net­works has reached the tar­get state, fur­ther neu­tral evo­lu­tion can lead to an increase in both the muta­tional robust­ness and the envi­ron­men­tal robust­ness to changes in ini­tial conditions.”

“The Pre­dic­tion API allows you to get more from your data and makes its pat­terns more acces­si­ble. Specif­i­cally, the Pre­dic­tion API lever­ages Google’s machine learn­ing infra­struc­ture to give you the tools to bet­ter ana­lyze your data and reveal pat­terns that are often dif­fi­cult to man­u­ally dis­cover. The API also enables you to use those pat­terns to pre­dict new out­comes, which facil­i­tates the devel­op­ment of all types of soft­ware, from tex­tual analy­sis sys­tems to rec­om­men­da­tion sys­tems. Because the Pre­dic­tion API is a REST­ful HTTP ser­vice, you can eas­ily access it from Google App Engine, Apps Script, and other Internet-​​connected desk­top applications.”

“Het­ero­gene­ity in the degree dis­tri­b­u­tion is known to sup­press global syn­chro­niza­tion in com­plex net­works of sym­met­ri­cally cou­pled oscil­la­tors. Scale-​​free net­works present strong het­ero­gene­ity, there are a few highly con­nected nodes, termed hubs, while the major­ity of nodes has only a few con­nec­tions. We show that a sta­ble par­tially syn­chro­nized state may take place in scale-​​free net­works: hubs undergo a tran­si­tion to syn­chro­niza­tion while the remain­ing nodes are unsyn­chro­nized. This phe­nom­e­non may occur in any large het­ero­ge­neous net­work, regard­less of the net­work global syn­chro­niza­tion prop­er­ties. We pro­vide the­ory and numer­i­cal evi­dence to estab­lish this phenomenon.”

“… In this paper we present BRACE (Big Red Agent-​​based Com­pu­ta­tion Engine), which extends the MapRe­duce frame­work to process these sim­u­la­tions effi­ciently across a clus­ter. We can lever­age spa­tial local­ity to treat behav­ioral sim­u­la­tions as iter­ated spa­tial joins and greatly reduce the com­mu­ni­ca­tion between nodes. In our exper­i­ments we achieve nearly lin­ear scale-​​up on sev­eral real­is­tic simulations.…”

“A wide fam­ily of non­lin­ear sequence gen­er­a­tors, the so-​​called clock-​​controlled shrink­ing gen­er­a­tors, has been ana­lyzed and iden­ti­fied with a sub­set of lin­ear cel­lu­lar automata. The algo­rithm that con­verts the given gen­er­a­tor into a lin­ear model based on automata is very sim­ple and can be applied in a range of prac­ti­cal inter­est. Due to the lin­ear­ity of these automata as well as the char­ac­ter­is­tics of this class of gen­er­a­tors, a crypt­an­a­lytic approach can be pro­posed. Lin­ear cel­lu­lar struc­tures eas­ily model keystream gen­er­a­tors with appli­ca­tion in stream cipher cryptography.”

“Con­nec­tions in com­plex net­works are inher­ently fluc­tu­at­ing over time and exhibit more dimen­sion­al­ity than analy­sis based on stan­dard sta­tic graph mea­sures can cap­ture. Here, we intro­duce the con­cepts of tem­po­ral paths and dis­tance in time-​​varying graphs. We define as tem­po­ral small world a time-​​varying graph in which the links are highly clus­tered in time, yet the nodes are at small aver­age tem­po­ral dis­tances. We explore the small-​​world behav­ior in syn­thetic time-​​varying net­works of mobile agents, and in real social and bio­log­i­cal time-​​varying systems.”

“The inverse Ising prob­lem is a dif­fi­cult com­bi­na­to­r­ial opti­miza­tion prob­lem in the class known as “NP-​​hard”. In the­ory, only approx­i­mate schemes, or meth­ods that take more than poly­no­mial time to find the answer are pos­si­ble. Boltz­mann Learn­ing [1] is an iter­a­tive method where in one step the cor­re­la­tion func­tions are com­puted given an Ising model, and in another step the Ising model cou­plings are mod­i­fied to adjust to data. In prin­ci­ple, Boltz­mann learn­ing can be employed to find the cou­plings with arbi– trary accu­racy given accu­rate data and suf­fi­cient time, but the slow con­ver­gence of the Boltz­mann learn­ing makes it a very inef­fi­cient algo­rithm for most prac­ti­cal purposes.”

“Approx­i­mate dynamic pro­gram­ming has been used suc­cess­fully in a large vari­ety of domains, but it relies on a small set of pro­vided approx­i­ma­tion fea­tures to cal­cu­late solu­tions reli­ably. Large and rich sets of fea­tures can cause exist­ing algo­rithms to over­fit because of a lim­ited num­ber of sam­ples. We address this short­com­ing using $L_​1$ reg­u­lar­iza­tion in approx­i­mate lin­ear pro­gram­ming. Because the pro­posed method can auto­mat­i­cally select the appro­pri­ate rich­ness of fea­tures, its per­for­mance does not degrade with an increas­ing num­ber of fea­tures. These results rely on new and stronger sam­pling bounds for reg­u­lar­ized approx­i­mate lin­ear pro­grams. We also pro­pose a com­pu­ta­tion­ally effi­cient homo­topy method. The empir­i­cal eval­u­a­tion of the approach shows that the pro­posed method per­forms well on sim­ple MDPs and stan­dard bench­mark problems.”

“There seems to be a brief recog­ni­tion period for Bdellovib­rio to iden­tify its prey after a col­li­sion with another cell (Shilo, 1969). Ini­tially, the attach­ment to a cell sur­face is reversible. Bdellovib­rio is still able to swim away a few sec­onds after rec­og­niz­ing that the cell is not a right tar­get (gram– pos­i­tive bac­te­ria). When a gram-​​negative bac­terium is encoun­tered, Bdellovib­rio cell becomes com­mit­ted to inva­sion. The whole process usu­ally takes around 5 – 10 min­utes. Bdellovib­rio drops its fla­gel­lum. It has been hypoth­e­sized that Bdellovib­rio may adhere to the cell sur­face using pilus-​​like fibre struc­ture expressed on its pen­e­tra­tion pole.”

“Small world net­works inter­po­late between fully reg­u­lar and fully ran­dom topolo­gies and simul­ta­ne­ously exhibit large local clus­ter­ing as well as short aver­age path length. Small world topol­ogy has there­fore been sug­gested to sup­port net­work syn­chro­niza­tion. Here we study the asymp­totic speed of syn­chro­niza­tion of cou­pled oscil­la­tors in depen­dence on the degree of ran­dom­ness of their inter­ac­tion topol­ogy in gen­er­al­ized Watts-​​Strogatz ensem­bles. We find that net­works with fixed in-​​degree syn­chro­nize faster the more ran­dom they are, with small worlds just appear­ing as an inter­me­di­ate case. For any generic net­work ensem­ble, if syn­chro­niza­tion speed is at all extremal at inter­me­di­ate ran­dom­ness, it is slow­est in the small world regime. This phe­nom­e­non occurs for var­i­ous types of oscil­la­tors, intrin­sic dynam­ics and cou­pling schemes.”

“Empir­i­cal stud­ies show that real world net­works often exhibit mul­ti­ple scales of topo­log­i­cal descrip­tions. How­ever, it is still an open prob­lem how to iden­tify the intrin­sic mul­ti­ple scales of net­works. In this arti­cle, we con­sider detect­ing the multi-​​scale com­mu­nity struc­ture of net­works from the per­spec­tive of dimen­sion reduc­tion. …The­o­ret­i­cal analy­sis indi­cates that all these three trans­for­ma­tions are cru­cial to iden­tify the multi-​​scale com­mu­nity struc­ture of net­works. Exten­sive tests on real world and arti­fi­cial net­works demon­strate that the cor­re­la­tion matrix sig­nif­i­cantly out­per­forms the mod­u­lar­ity matrix as regards iden­ti­fy­ing the multi-​​scale com­mu­nity struc­ture of networks.”

“Now, to return to the news arti­cle. If the eigen­value dis­tri­b­u­tion is an attrac­tor, this means that a lot of phys­i­cal and social phe­nom­ena which can be mod­eled by eigen­val­ues (includ­ing, appar­ently, quan­tum energy lev­els and some prop­er­ties of sta­tis­ti­cal tests) might have a com­mon struc­ture. Just as, at a sim­i­lar level, we see the nor­mal dis­tri­b­u­tion and related func­tions in all sorts of unusual places.”

“…Through sev­eral nu– mer­i­cal exper­i­ments, we demon­strate that both matrix– and tensor-​​based tech­niques are effec­tive for tem­po­ral link pre­dic­tion despite the inher­ent dif­fi­culty of the prob­lem. Addi­tion­ally, we show that tensor-​​based tech­niques are par­tic­u­larly effec­tive for tem­po­ral data with vary­ing peri­odic patterns.”

“I envi­sion a future in which pro­gram­mers are the con­scious repos­i­to­ries of a body of knowl­edge. A future in which they regain their craft, instead of tweak­ing frame­works they don’t under­stand. A future, even­tu­ally, in which pro­gram­mers say “no” to demands at odds with their ethics.

It is cru­cial to cre­ate ways, spaces and for­mats for pro­gram­mers to share their knowl­edge with other pro­gram­mers. It is vital we keep this knowl­edge (espe­cially ver­bal­ized knowl­edge) among pro­gram­mers and out of salespeople’s hands. And it is urgent the IT crowd rec­og­nize soft­ware mak­ing as a craft, instead of a commodity.”

“We con­sider sched­ul­ing weighted pack­ets in a capacity-​​bounded buffer. In this model, there is a buffer with a lim­ited capac­ity B such that at any time, the buffer can­not accom­mo­date more than B pack­ets. Pack­ets arrive over time. Each packet has a non-​​negative real value. Pack­ets do not expire and they leave the buffer only because either we send them or we drop them. The pack­ets that have left the buffer will not be recon­sid­ered for deliv­ery any more. In each time step, at most one packet in the buffer can be sent. The order in which the pack­ets are sent should com­ply with the order of their arriv­ing time. The objec­tive is to max­i­mize the total value of the pack­ets sent in an online manner.…”

“Each com­bi­na­tion of val­ues was then fed into a sim­u­la­tor to give an idea of the grid’s per­for­mance and its expected life­time. If the per­for­mance was promis­ing, the “genetic mate­r­ial” was sub­jected to fur­ther ran­dom changes, or muta­tion, and this process was repeated until no more improve­ments were forthcoming.

After 100 gen­er­a­tions, the GA spawned a geometry/​voltage set that boosted the ion engine grid’s life­time to 5.1 years — at least in the sim­u­la­tor (Jour­nal of Propul­sion and Power, DOI: 10.2514÷1.44358). Fac­tors opti­mised included grid hole diam­e­ter, hole spac­ing and the thick­ness of the grids. The engine could be improved fur­ther, says Far­nell, by evolv­ing the other parts too.…”

“Results on a com­modi­ties basket”

“In this paper we con­sider spa­tial net­works that real­ize a bal­ance between an infra­struc­ture cost (the cost of wire needed to con­nect the net­work in space) and com­mu­ni­ca­tion effi­ciency, mea­sured by aver­age short­est path­length. A global opti­miza­tion pro­ce­dure yields net­work topolo­gies in which this bal­ance is opti­mized. These are com­pared with net­work topolo­gies gen­er­ated by a com­pet­i­tive process in which each node strives to opti­mize its own cost-​​communication bal­ance. Three phases are observed in glob­ally opti­mal con­fig­u­ra­tions for dif­fer­ent cost-​​communication trade-​​offs: (i) reg­u­lar small worlds, (ii) star-​​like net­works and (iii) trees with a cen­tre of inter­con­nected hubs. In the lat­ter regime, i.e. for very expen­sive wire, power laws in the link length dis­tri­b­u­tions $P(w)\propto w^{-\alpha}$ are found, which can be explained by a hier­ar­chi­cal orga­ni­za­tion of the networks…”

“We present a near-​​linear time algo­rithm that approx­i­mates the edit dis­tance between two strings within a poly­log­a­rith­mic fac­tor; specif­i­cally, for strings of length n and every fixed epsilon>0, it can com­pute a (log n)^O(1/epsilon) approx­i­ma­tion in n^(1+epsilon) time. This is an expo­nen­tial improve­ment over the pre­vi­ously known fac­tor, 2^(O (sqrt(log n))), with a com­pa­ra­ble run­ning time (Ostro­vsky and Rabani J.ACM 2007; Andoni and Onak STOC 2009). Pre­vi­ously, no effi­cient poly­log­a­rith­mic approx­i­ma­tion algo­rithm was known for any com­pu­ta­tional task involv­ing edit dis­tance (e.g., near­est neigh­bor search or sketching). ”

“In many dynam­i­cal sys­tems there is a large sep­a­ra­tion of time scales between typ­i­cal events and “rare” events which can be the cases of inter­est. Rare-​​event rates are quite dif­fi­cult to com­pute numer­i­cally, but they are of con­sid­er­able prac­ti­cal impor­tance in many fields: for exam­ple tran­si­tion times in chem­i­cal physics and extinc­tion times in epi­demi­ol­ogy can be very long, but are quite impor­tant. We present a very fast numer­i­cal tech­nique that can be used to find long tran­si­tion times (very small rates) in low-​​dimensional sys­tems, even if they lack detailed bal­ance. We illus­trate the method for a bistable non-​​equilibrium sys­tem intro­duced by Maier and Stein and a two-​​dimensional (in para­me­ter space) epi­demi­ol­ogy model.”

“Image Effects With Cel­lu­lar Automata (PDF) Abstract:This paper presents some tech­niques for cre­at­ing var­i­ous artis­tic effects on dig­i­tal pho­tog­ra­phy using the con­cept of cel­lu­lar automata. All exam­ples in this paper are cre­ated by “Image Infec­tor” pro­gram, which is a plu­gin for Pixo­pe­dia 24 image edi­tor and painter (www​.sigmapi​-design​.com).”

“I’ll start by high­light­ing that while all the soft­ware in this post is indeed free (true to FOSS), an account with Inter­ac­tive Bro­kers is needed to make use of it. For those not famil­iar with IB, they offer a trad­ing plat­form that excels on numer­ous fronts but is most appeal­ing to those of us who trade algo­rith­mi­cally. IB makes avail­able a rather com­pre­hen­sive API that makes data access and trade exe­cu­tion entirely pos­si­ble pro­gram­mat­i­cally via a hand­ful of “sup­ported” lan­guages. These include Java (the lan­guage of the plat­form), C#, VBA and even Excel. The also have a POSIX com­pli­ant C++ ver­sion for those who enjoy C++ but dis­like Windows.

For those who dis­like Win­dows and C++, the com­mu­nity of IB users have a few “non-​​official” options. They include some nice imple­men­ta­tions in C, Python (2), Mat­lab, and some­thing even more abstracted in the trading-​​shim. While all well and good, there was one miss­ing: R.…”

“This paper aims at gen­er­at­ing a new face based on the human like descrip­tion using a new con­cept. The FASY (FAce SYn­the­sis) Sys­tem is a Face Data­base Retrieval and new Face gen­er­a­tion Sys­tem that is under devel­op­ment. One of its main fea­tures is the gen­er­a­tion of the requested face when it is not found in the exist­ing data­base, which allows a con­tin­u­ous grow­ing of the data­base also.”

“The per­for­mance assess­ment of algo­rithms for mul­ti­ob­jec­tive opti­miza­tion prob­lems is far from being a triv­ial issue. Recent results indi­cate that unary per­for­mance mea­sures, i.e. per­for­mance mea­sures which assign a sin­gle value to each non-​​dominated point set, are inher­ently lim­ited in their infer­en­tial power. Despite these lim­i­ta­tions, the hyper­vol­ume indi­ca­tor (also known as Lebesgue mea­sure or S met­ric) is still con­sid­ered to pos­sess some rea­son­able prop­er­ties, hav­ing also been pro­posed as a guid­ance cri­te­rion for accept­ing solu­tions in Mul­ti­ob­jec­tive Evo­lu­tion­ary Algo­rithms. There­fore, the com­pu­ta­tional time taken for com­put­ing the hyper­vol­ume indi­ca­tor is a cru­cial fac­tor for the per­for­mance of such algorithms.…”

“Unfor­tu­nately, our polit­i­cal lead­ers don’t give a damn about mun­dane issues such as unem­ploy­ment and eco­nomic growth. It is far eas­ier for them to bandy about silly cliches about fis­cal respon­si­bil­ity and gen­er­a­tional equity, even though the poli­cies they are push­ing are 180 degrees at odds with any­thing that will help our chil­dren or grand­chil­dren. Their main con­cern is push­ing poli­cies that keep the finan­cial indus­try happy. And 10 mil­lion unem­ployed never both­ered any­one at Gold­man Sachs, just as Fab­u­lous Fabio.”

“We believe that our result could shed the light on the rela­tion between the dis­tri­b­u­tion of motifs and the struc– ture of the adja­cency matrix of a hier­ar­chi­cal net­work. How­ever to make this rela­tion more pro­found the “in– verse” prob­lem should be con­sid­ered as well. Namely, it would be desir­able to check if the sta­ble dis­tri­b­u­tion of motifs is uniquely related to any kind of hier­ar­chi­cal orga­ni­za­tion of the network.”