Release Notes 3.24.3 June 30 2010 Changes -avg cell size, and avg individual's cell size Bug fixes -avg degree was made double Release Notes 3.24.2 May 6 2010 Changes - corrected copyright notice Release Notes 3.24.1 Jan 18 2010 Changes Bug fixes -link set dirty is now correctly handled Release Notes 3.24 Jan 18 2010 Changes -testing of all metrics, and most of the functionality -agreement is redefined by removing its normalization (computational efficiency) -documentation of all parameters -more efficient & accurate code for tracking worstFriendAgreement * agreement is stored in each link; nodes find the min over the links * agreement at a link is updated lazily: only when asked Bug fixes -error in assortativity and E-I_index formulae -error in randomness of pressure Release Notes 3.23 Jan 11 2010 Changes -more complete reporting of summary statistics -network display starts sorted -E-I index -more metrics on components: number of radical components, and number of isolated radical components -changes to names of most metrics (typically now "R-something" to reference radicals) -comprehensive documentation of all metrics Bug fixes -number of ticks is now exactly runLength variable (before there was an extra tick) -soffer's metric now estimates degree using the whole graph, not just the radical subgraph -fixed problems loading unit test file Release Notes 3.22 Jan 8 2010 Changes -switched magnets to HashMap for performance -better handling of errors -unit test file Bug fixes -correct formula for computing clustering with Watts-Strogatz formula -error computation in computation of median of radical neighborhood size Release Notes 3.21.1 Jan 7 2010 Changes -cleanup of obsolete features, some better documentation -reporting of fraction of rad-rad links -removed radicalExcessDyads Bug fixes -output files are now closed on exit Release Notes 3.21 Jan 7 2010 Changes -added cell size metric to display -added reporting of traditional clustering measure -added clustering data for all nodes (not just radicals) -moved label counter -cleanup of interface for pair data and triangle data -tester of isolation and clustering metrics Bug fixes -many fixes to the triangle data computation -incorrect reporting or initIsolation and finalIsolation -clustering was reported with a fixed factor of 3 error -median cell size was not the component size but rather number of radical neighbors of radicals Release Notes 3.20.1 Dec 28, 2009 Bug fixes -changed algorithm for how magnets form links (no sociological difference) apparently resolving heap explosion; should also create speedup. -energy values did not CORRECTLY report linkAgeBonus effect Release Notes 3.19.1 Dec 7, 2009 Changes -reporting of full network data Release Notes 3.18 Nov 15, 2009 Changes -new output of raw network (of radicals only) -new parameter: networkRecordTic. special value -1 indicates print at end of simulation. Release Notes 3.17 Oct 23, 2009 Changes -internally nicer handling of statistics -api for testing -test scripts Bug fixes -energy values did not report linkAgeBonus effect -possible bug in forming new ties -missing comma in output (REM 3.17.1) Release Notes 3.16 Changes -most parameters are now stored in static variables, leading to (hopefully) a much smaller memory footprint. -meanRadicalComponentSize is now computed -magnets' lists are now TreeSets rather than ArrayLists: faster removal and avoidance of duplication Bug fixes -crash in doing LHS -error in path of raw files -better handling of different modes Release Notes 3.15 Changes -transitive ties are now a continuous trait: 0..1, with a default of 0.5 -caching of friendship agreement -mutual friendship factor functionality disabled: considerable complications with caching of friendship agreement unresolved issues with oscillations due to introductions -ethnicGpOne has been replaced by "diversity" which runs 0..1 and affects all attributes except for zeal. -new parameters: fixedAttribSalience and zealSalience (range: 0..10, default=2. both) multiply the importance of those node properties. -added histeresis to the model by giving ties a bonus that's related to their age. bonus = maxLinkAgeBonus * (linkAge)/(linkBonusConstant + linkAge) Bug fixes -the failure of the attribVar batch to work (duplicated runs at numVarAttribs=1) is due to the batch running the =0, but the program automatically moving it to the threshold of 1. This threshold was lowered to 0. Similarly for fixed attributes. Release Notes 3.14 Changes -mutual friends now affect tie formation, energy (TODO: update writeup) -new parameter "mutualFriendFactor" affects the weight of a mutual friend in the computation of agreement (TODO: update writeup) -new median cell size metric -update to how aggreement is computed with nodes of zeal=0 (TODO: update writeup) -defaults changed: fixedAttributes->5, Magnetic Encounter Rate->.2 -exposed all parameters: many new parameters now introduced -reverted to average degree -renamed "peerPressureRate" to "avgPressurability" -changed attrition rate to 2E-3/tick = 10%/year (at 50 ticks/year) Release Notes 3.13 Changes: -Degree of node is now distributed as a power law with default exponent, nodeDegreeExponent=2.25, giving an average degree of 5. -New metrics -Rewamped random number system -Cleanup -Initial degree of the node is now stored permanently. -New friends are allowed without loss of old ones as long as there is a void under the cap -Friendships are now distributed with a power law, with default exponent of 2.25 (avg=5) Bug fixes: -As per the paper, enforce the policy that no friends would be allowed to tie in if there already have maxDegree friends. Release Notes 3.12 Changes: -Remove a call that prevented nodes from having < 1 fixed attributes. Release Notes 3.11 Changes: -parameter values entered into the panel before reset is clicked are now saved. -parameter values have been changed to their values in the paper (except population size) -renamed metrics for consistency with paper Bug fixes -small bug involving the UI of radicalAttritionIncrement -fixed a bug that prevented execution from the command line -fixed a bug with cached random numbers -fixed the problem in displaying node labels. -fixed a bug corrupting the system.out stats Known Issues 1. Opening a node panel, and then clicking X freezes up the program, rather than quitting. The bug does not show in debug mode! 2. Having a variable parameter in a batch file accidentally set to a fixed value elsewhere in the batch file causes the program to go into an infinite loop, perpetually iterating the first seed and the fixed value. 3. Memory consumption: currently, running a 50,000 agent simulation requires 2GB of memory, primarily because random number are cached at each node. This can be drastically reduced by setting randomCacheSize to a smaller number (now it's 200). The advantage of the current architecture is that it allows the simulation at startup to be somewhat insensitive to model parameter values. In the future, we could make the nodes drain their cache after the model started up. Release Notes 3.10 Changes: -parameter numMagnets has been renamed numMagnetsN (N for neutral, for clarity) -parameter numFixedTraits has been renamed numFixedAttributes -parameter numTraits has been removed. Instead, there is a parameter numVarAttributes -new sequences display the clustering of radicals, and the tie strengths in radicals. -it is now possible to click a node, and the "probe pabel" would change the values of the tie strengths each tick. Previously, it would display out-of-date information -creation of new networks now uses caching of random seeds. This ensures that changes to parameter values like number of magnets has no effect on the states of the nodes and their connections. Thus, the repeatability of instances is increased. -the network display (in circle layout) now shows all the radicals on one part of the screen, then all the moderates, and finally all the pacifists. Bug fixes: -outputDirectory is created if it is not existent -edge creation process is now fair: previously, nodes with low numbers tended to have a low number of edges. -fixed a bug preventing radicalAttritionIncrement from functioning properly. Release Notes 3.9 Changes: -better statistical reporting: * at each turn, the simulation writes vital radicalization indicators to a .csv file under the temporary files directory * at the end of a run, the average values of vital statistics are reported, not just the initial and final values. -it is now possible to disable influence between nodes by setting peerInfluenceRate << 0.0, eg. -1.0 -3 new metrics of radicalization: avgRadicalTieStrength = how strong is the average tie between two radicals radicalTriads = fraction of triads connecting three radicals, out of possible triads radicalClustering = same metric, except that we look at the strength of the ties in each triad, and calculate the average tie strength over the three ties (radicalTriads metric is equivalent to assuming it's = 1) -fractionInitPacifists is now 0.05, the same as the number of radicals, to remove bias. -peerPressureRate is now much lower at 0.2 (previously=0.7). Bug fixes: * serious bug that was causing all ties to be +/-1.0 * small bug related to handling of UNIX-based systems. * solved most compilation warnings * there were two serious errors in how peer pressure was calculated b/c of me misinterpreting Macy's paper. Release Notes 3.8 Changes: -endogenized magnets: during creation, each node randomly selects whether to visit one of the magnets. This preference is specific to each of the magnets. This creates additional realism to the simulation, * magnet is an individual preference, specific to the activity, rather than just an on/off switch * magnet is now a trait in the Hopfield network sense, alongside other traits, and not an exogenous rewiring action. The chance of selection of each of the magnets is dependent on the fractionExposedToMagnets parameter (needs a better name). For example, if there is a radical node and there are 10 magnets, 5 radical magnets, and 4 pacifist magnets, the node has probability fractionExposedToMagnets to enroll in magnet 1, likewise, the same independent probability of enrolling in magnet 3 .. 10, and then similarly for the radical magnets. If would not enroll in the pacifist magnets. Node that becomes a radical/pacifist would look at the locales frequented by radicals, and may enroll, or not, depending on the fraction encounter rate. Node that ceases to be a radical would cease to attend radical magnets. In the above, all that was said of radical equally applies to pacifists and moderates. Bug fixes: -suspected crash when attrition is high, but the network display is not open. Release Notes 3.7 Changes: -by clicking on nodes, it's now possible to see useful information about them and their links -radical attrition rate increment can now go to negative values -changing model parameter will now immediately update the control panel, displaying any changes to parameters that are linked to the one being changed. For example, setting fracInitRadicals to .90 would show a decrease in fracInitPacifists to .10 (if it was > 0.1 before. This is necessary to ensure that the sum of the two numbers cannot exceed 100%). Bug fixes: -at numZealLevels = 2, 4, moderates will now NOT be created. Release Notes 3.6 Changes: -parameter fractionInitModerates has been removed, and replaced with the parameter fractionInitPacifists. This is because moderates do not actually exist if number of levels is even, but radicals always do. -When the number of levels is or set to be an even value (2,4,6 etc), automatically, we set: fractionInitPacifists = 1 - fractionInitRadicals Release Notes 3.5 Changes: -renamed "plot" and "plot2" parameters to more informative names: "showEnergyPlot", "ShowStatsPlot" -both clerics have been disabled. -agents can now join their friends, much like magnets do. "In transitivity" -each agent picks two of his friends every tick and tries to join them. -this can be disabled by a new parameter, "transitiveFriendship" -safeMagnets: renamed "safeNewFriends". -it now affects not only magnets but also transitivity of friendship -numDisciples: renamed "population" -attrition: -new parameters: "attritionRate" (0..1), "radicalAttritionIncrement" (0..1, such that common+radicalInc <= 1) -a fraction commonAttrition of the non-radical population is removed each term. -a fraction common+radicalAttritionIncrement of radicals is removed. -new individuals who number exactly as many as the lost ones are brought in. Their statistical characteristics are given by FractionInitModerates & FractionInitRadicasl - the same as the for the initial model population. They form random new ties among themselves and the general population. -runLength: the time till termination will now be 100 be default. Changing the number of agents will not affect this. Please set your run length to be sufficiently large if you are running a large social network. -bug fixes: -a serious problem affecting numZealLevels! -a serious problem affecting magnets!