Human dual task datasets (networks data, ATTENTION! Large file: 2.2 Gb)

Complete human dual task datasets (7 Gb) including all experiments with activity (phase and amplitude, see Sigman paper below for details) of each voxel, code to read the data, spatial coordinates of the nodes, and activity mask used to filter the data.

Human resting state datasets (network datasets, ATTENTION! Large file: 1.4 Gb)

These files contain all networks information and readme.txt files for networks of networks in dual tasks and resting states in functional human brain networks.

For more details on the datasets, please see here.

Repository of all gaze position data and code

The above repository contains all the data and code used for our work on applying a Maximum Entropy model to gaze positions recorded from a set of subjects who were instructed to watch and rate Super Bowl 2014 commercials. The individuals’ gaze data, demographic data, and ratings are included, as well as some ratings data at the larger scale of the general United States public.

There is also a readme.txt explaining the meaning and functions of all the code and data.

2014 Super Bowl commercials viewed by the participants in the study. Gaze positions for all viewers are denoted by red dots.

Please find here the code (in C) for the algorithm used in the paper of Pei, Teng, Shaman, Morone, and Makse, Efficient collective influence maximization in cascading processes with first-order transitions, Sci. Rep. 7, 45240 (2017).

This code calculates values of Collective Influence in the Threshold Model (CI-TM) in an Erdos-Renyi random graph; the structure of the graph is a list of neighbors.

Please find here the dataset of Twitter retweets used in the paper of Morone and Makse, Influence maximization in complex networks through optimal percolation, Nature 524, 65–68 (2015).

Algorithm to identify superspreaders using Collective Influencers (CI), k-shell, and PageRank.

Please find CI algorithm and code. CI algorithm is explained in F. Morone, H. A. Makse: Influence maximization in complex networks through optimal percolation, Nature (2015).

Instructions to run the CI code are as follows:

• to compile the source code, use this command: gcc -o CI CI_HEAP.c -lm -O3
• inputs are the file containing the network, and your desired value of L (the radius of the ball you want to consider)
  • network file must be an adjacency list where the first number on each line is a node ID and the following numbers on that line are its neighbors, e.g.:
    1 3255 18210 24119 70247
    2 9205 88665 89859
    3 3255 328244 25046 41508
    …
    3255 1 3 34913 73168
• to run the code, use this command: ./CI <network_filename> <L>
• values of (q,G) are printed on the screen for fraction of removed nodes q, and an output file “Influencers.txt” is created listing the influencers in the network by decreasing influence

Algorithms of k-shell and pagerank are in this link. More information is in S. Pei, L. Muchnik, J. S. Andrade, Jr., Z. Zheng, H. A. Makse, Searching for superspreaders of information in real-world social media, Sci. Rep. 4, 5547 (2014).

For detailed explanation see here.

The datasets were published in S. Pei, L. Muchnik, J. S. Andrade, Jr., Z. Zheng, H. A. Makse, Searching for superspreaders of information in real-world social media, Sci. Rep. 4, 5547 (2014). DOI: 10.1038/srep05547. [PDF].

Visualization:

The cover of our Nature Physics 2010 paper on kcore spreaders M. Kitsak, L. K. Gallos, S. Havlin, F. Liljeros, L. Muchnik, H. Eugene Stanley, H. A. Makse, Identification of influential spreaders in complex networks, Nat. Phys. 6, 888 (2010). [PDF]

Cover:

was created by the software package of LaNet-vi, which you can download at LaNet-vi. It was developed by the group of Indiana U. There is also a webpage interface you can use there. You may want to try different parameters to adjust the figures as you wish.

Beside this figure, most of our Nature-like high-quality figures are done with MetaPost. For instance, Fig. 1 in S. Pei, et al. Sci. Rep. 4, 5547 (2014). PDF:

Unfortunately, this is not an ‘interactive’ software to generate these type of images, but it is a programmable description of the image. If you google metapost tutorial you will find a lot of pages explaining the details. In general, it is not too hard but you would need to spend some time to learn how it works.

In general, a metapost file includes commands such as: fill fullcircle scaled 8.0 shifted (253,385) withcolor (1.0,0.0,0.0); which draws a circle of radius 8.0 at coordinates 253,385, using the red color, which in rgb is (1,0,0). You need to write a program that will take as input the nodes coordinates and radii, as well as their color, and then produce a file with statements such as the above. This metapost file needs to start with beginfig(1); and end with endfig; end;

This means that a file, test.mp , may look like beginfig(1); fill fullcircle scaled 8.0 shifted (253,385) withcolor (1.0,0.0,0.0); endfig; end;

You then need to compile this file with the command mpost test.mp This will produce a file called test.1 which is a postscript file that you can rename to test.eps and is your final image.

Before preparing the mp file, you may need to use other network visualization softwares to get the coordinates of nodes, such as gephi and pajek. Then you can export the coordinates and incorporate them into the mp file. We attach two mp files as examples. You can write your own mp files based on these examples: twolayer_c1.mp , and twolayer_sample.mp .

The datafiles used in the Srep paper are:

• APS Dataset This file contains the coauthorship and citations of all scientific papers published in American Physical Society (APS) journals until 2005, including Physical Review A, B, C, D, E and Physical Review Letters.
• Facebook Dataset This dataset is available online at http://socialnetworks.mpi-sws.org/data-wosn2009.html. It contains the friend relations of New Orleans Facebook social network as well as the wall posts records of users during a period of nearly two years. In the social network there are 63731 nodes with average degree 24.3. The total number of wall posts is 876992.
• Twitter Dataset This file contains the mention network and retweet relations extracted from the tweets sampled between January 23rd and February 8th, 2011 provided by Twitter (http://trec.nist.gov/data/tweets/). We are not allowed to distribute any private information about the twitter users, so in the dataset each user is represented by an anonymized ID. Retweet dataset in Twitter that we used in a collective influence paper can be downloaded here.

For analysis you will need the following:

1. Gephi (for visual analysis)
2. Anaconda for Python 3.6
3. The Twitter account you created in Step 1
4. Python files necessary for Twitter opinion mining

For full instructions as well as a tutorial, please see the separate Twitter Analysis page in our website.

For those who would like to learn the R or Python computer languages from scratch, please see the following compilation of resources.

Please find the homepage of Kcore Analytics here: http://www.kcore-analytics.com/

Employing state-of- the-art theories of combinatorial optimization, we have developed a framework to solve the influence maximization problem in complex networks, along with a new paradigm of optimal collective influence. This seminal work has been published in: Morone & Makse, Influence maximization in complex networks through optimal percolation, Nature 524, 65-68 (2015).

We have embodied our algorithm as a large-scale search engine, named Collective Influence (CI), for localizing the most influential individuals in complex social networks, like Twitter. The service is currently patent pending and is available online at the address: http://www.kcore-analytics.com.

CI identifies an optimal set of nodes, called influencers, which may potentially spread information in the entire network in the fastest possible way. The CI algorithm is highly scalable to present day big-data. This method identifies the most influential people by introducing a ranking method based on collective behavior of people in a social network and identifies the minimal set of such people that are
able to spread information throughout the whole network.

Using the search engine at http://www.kcore-analytics.com is very simple. The user just needs to type the topic of interest in the search bar (for example a keyword like big data, global warming, oil…). Then, the algorithm collects all Tweets mentioning the given topic and reconstructs the network of interactions among users. Finally, CI identifies the most influential users, which are displayed according to their Collective Influence ranking.

In our recent work, we applied our recent findings regarding the collapse of dynamical complex systems to the field of ecology and finance. Our main result is that these systems show a remarkable similarity in the ways that species or stocks populate their given networka. Precisely, both ecological and financial networks show a very high density of species/stocks at the periphery, as well as in the innermost core of the network. On the contrary, the network region separating the core from the periphery is rather diluted and scarcely populated. Based on this commonality in the network architecture, we further speculate that collapses of ecological and financial systems can be explained in a unified way through the core-periphery organization of their underlying networks.

• IMDB Dataset (AdultIMDB.dat.gz): IMDB actors in adult films. We have created a network of connections between actors who have co-starred in films, whose genre has been labeled by the Internet Movie Database as ‘adult’. This network is a largely isolated sub-set of the original actor collaboration network. Additionally, all these films have been produced during the last few decades, rendering the network more focused in time. The largest component comprises 47719 actors/actresses in 39397 films. The average degree of the network is 46.0.
• Email Dataset (Emailcontacts.dat.gz): The network of email contacts is based on email messages sent and received at the Computer Sciences Department of London’s Global University. The data have been collected in the time window between December 2006 and May 2007. Nodes in the network represent email accounts. We connect two email accounts with an undirected link in the case where at least two emails have been exchanged between the accounts (at least one email in each direction). There are 12701 nodes with an average degree of 3.2.
• POK Dataset (pok.dat):This file contains the entire sequence of messages sent by users in the POK community. The three columns correspond to: (sending member ID) (receiving member ID) (seconds). The third column shows the time when the message was sent, measured in seconds since the very first message.

We have developed a novel filtering approach based on percolation analysis to construct the fractal structures of the present-day protein-protein interaction (PPI) networks. The original unfiltered data of present-day networks are downloaded from the STRING database.

Ancestral proteins are reconstructed using orthologous groups on different evolutionary levels. The information of orthologous groups are obtained from the eggNOG database. Ancestral interactions are reconstructed using a stochastic duplication-divergence model. For more information, please refer to our paper. All files below are .rar

• Dataset for the PPI networks of seven species (E. coli, S. cerevisiae, A. thaliana, C. elegans, D. melanogaster, M. musculus, H. sapiens) at different evolutionary levels.
• Algorithm and simple instructions to perform the percolation analysis.
• Algorithm and simple instructions to perform the static analysis (fractal dimension, degree distribution, renormalization group analysis…).
• Algorithm and simple instructions to perform the dynamic analysis (test of the multiplicative growth mechanism).
• Mapping between ID of each node and the String identifier. File = mapping.txt.

Algorithm to perform a box covering to calculate the fractal dimension of a complex networks using the MEMB algorithm in C. All algorithms are explained in C. Song, L. K. Gallos, S. Havlin, H. A. Makse, “How to calculate the fractal dimension of a complex network- the box covering algorithm”, J. Stat. Mech. , P03006 (2007). See paper in JSTAT. See enclosed “readme.txt” file for instructions.

Algorithm to calculate the fractal dimension of a network. This file includes MEMB, CBB and random covering algorithms for Python. This package requires networkx. See file header for comments on how to use it.

• Download datasets we used in our research on the City Clustering Algorithm (CCA) and population growth. Here is the full dataset with readme instructions. Additionally, each datafile is below:
  • Dataset for USA in ascii format. Also available in Excel format here.
  • Dataset for GB for years 1981 and 1991.
  • Dataset for population of USA MSAs.
  • Dataset for boundaries of USA MSAs in 2000.
  • Dataset for population of USA Places in 2000.
  • Dataset regarding boundaries for Places or FIPS are too large. We make it available under request or they can be downloaded directly from the US Census Bureau.
• Download results obtained with the CCA algorithm in the USA and GB.
  • Download FIPS corresponding to each CCA cluster, using the CCA algorithm, for l=2,3,4,5km.
  • Download FIPS corresponding to each MSA.
  • Download MSA/CCA correspondence.

CO₂ Emissions Data

• Original dataset used in the Scientific Reports paper by Oliveira, Andrade, and Makse, Large cities are less green (2014).
  • Contains input (grid_xypci.dat) and output (out_CCA_d1000_l5.dat) data for total CO₂ emissions.
  • Input is arranged in columns: longitude, latitude, population, CO₂, income; output was found using parameters of population threshold D* = 1000 people/km² and distance threshold ℓ = 5 km.
  • Area of each cell is approximated by 1 km²; the calculation of this area is a difficult problem involving spherical triangles.
  • This dataset is on Lambert conformal conic projection, as opposed to the commonly-used Equirectangular projection.
• Global Rural-Urban Mapping Project (GRUMPv1) dataset; this data is already on the Equirectangular projection.
  • Data is arranged as follows:
    lx ly m lon_northwest lat_northwest (top row only)
    i   j   pop(i,j)
  • Image based on the above data.
• New CCA (City Clustering Algorithm) code with area calculation and associated GRUMPv1 dataset.
  • Density of each cell is defined as population/area; parameters used are D* = 1500 people/km² and ℓ = 5 km.
  • Dataset is already divided for each CCA city and is on the Equirectangular projection.
  • Data is arranged as follows:
    lx ly m lon_northwest lat_northwest (top row only)
    i   j   pop(i,j) area(i,j)
  • Images based on the above data for Los Angeles, New York City, and Chicago.

• These are the datasets we used in our Obesity research. See paper: Gallos, Barttfeld, Havlin, Sigman, Makse, “Collective behaviours in the spatial spreading of obesity”, Sci. Rep. 2: 454 (2012). These include data on population, Obesity, Diabetes, Food Services, Supermarkets, County Fips, Economy among others. Cancer data must be obtained directly from the website stated in the paper, since it is copyrighted.
  • Diabetes Dataset The fips file contains four columns in the following order: The first is the county ID assign by us arbitrarily, the second is the fips code of the county, the third is the state of the county and the last is the county name. these are organized alphabetically. There are 3092 counties in total, and all the files in the Census Research section contain the same number of counties.
  • population Dataset in this file the first column corresponds to the counties for years 1969 -2009. the remaining columns shoe the population for each subsequent year.
  • pop density Dataset same as above for population/ square mile.
  • Obesity Dataset Percentage of adult obese people (out of a 100) in each county for 2004-2008.
  • County fips Dataset Percentage of adult people with diabetes (out of a 100) in each county for 2004-2008.
  • Physical inactivity Dataset Percentage of adult people who are not physically active (out of a 100) in each county for 2004-2008.
  • Employment Dataset Fraction of people in a county who are Employed in any sectors of the economy. Years 1986 – 2009.
  • Food Service Employment Dataset Fraction of people employed in the food service industry. (NAICS code 722) Years 1986 – 2009.
  • Food stores employment Dataset Fraction of people employed in food stores. (NAICS code 445) Years 1986 – 2009.
  • Supermarket Employment Dataset Fraction of people employed in supermarkets. (NAICS code 44511) Years 1986 – 2009.
  • Wholesale Employment Dataset Fraction of people employed in Wholesale. (NAICS code 42) Years 1986 – 2009.
  • Administrative jobs Dataset Fraction of people in administrative jobs. (NAICS code 56) Years 1986 – 2009.
  • Manufacturing Employment Dataset Fraction of people employed in Manufacturing. (NAICS code 31) Years 1986 – 2009.

Algorithm and simple instructions for the protocol to prepare packings of spherical particles using Molecular Dynamics as usually done in our papers. This file includes the MD-Distinct Element Method (DEM) for Fortran. See the instruction manual to use the code.

• Download the datasets we used in our research on the phase diagram of jammed matter that predicts RCP. This is the data published in Song, Wang, Makse, Nature 453, 629 (2008). The data contains the box size of the system, the coordination for each particle in the system and the force between any two touched particles.
  • Dataset for RCP line in the phase diagram. The instruction for reading the data here.
  • Dataset for RLP line in the phase diagram. The instruction for reading the data here.
  • Dataset for G line in the phase diagram. The instruction for reading the data here.

• Codes. This zip-file contains the routines to calculate the Voronoi boundary and the contact radius for dimers and spherocylinders (as Mathematica notebook). It also contains code to calculate the excluded volume and surface for spherocylinders, lens-shaped particles (C-code) and dimers (Mathematica notebook). We have included the simulation code in F90 to generate packings of spherocylinders based on “Trubal” (see above under “Research on hard sphere packings”). For a detailed explanation of the Voronoi calculation for lenses and other shapes, see the report of Louis Portal.
• For more details on the algorithms used, please see the publication: Mean-field theory of random close packings of axisymmetric particles by A. Baule, R. Mari, L. Bo, L. Portal, H. A. Makse, Nature Communications 4, 2194 (2013).

Algorithm in Python to calculate the volume fraction of binary packings of hard spheres. Please cite the following paper if you use this code:

M. Danisch, Y. Jin, and H. A. Makse, Model of random packings of different size balls, Phys. Rev. E 81, 051303 (2010). [pdf].

The full data coordinates of all packings is here.

Algorithm and instructions to generate hard spheres packings at mechanical equilibrium from RLP to FCC. The computer simulation of jammed packing includes two independent algorithms. In the first part, the modified Lubachevsky-Stillinger (LS) algorithm is used to generate hard-sphere packings. This code was obtained from the website of Torquato at Princeton. Please cite the following papers (available at Torquato’s website) if you use this code: A. Donev, S. Torquato, and F. H. Stillinger, Neighbor List Collision-Driven Molecular Dynamics for Nonspherical Hard Particles: I. Algorithmic Details, Journal of Computational Physics, 202, 737 (2005). A. Donev, S. Torquato, and F. H. Stillinger, Neighbor List Collision-Driven Molecular Dynamics for Nonspherical Hard Particles: II. Applications to Ellipses and Ellipsoids, Journal of Computational Physics, 202, 765 (2005).

In the second part, the computer program “Trubal” applies MD or Distinct Element Method (DEM) to generate mechanically stable jammed packings of particles interacting with Hertz-Mindlin forces. TRUBAL was initially developed by Cundall and Strack and it is commonly known in the engineering literature as Discrete Element Method. Our version of the code is a more developed version based on the original Cundall code and that developed by Prof. Colin Thornton at the Department of Civil Engineering at Aston University. The original manuals with tons of explanations of the structure of DEM and the interactions considered between the particles can be downloaded here:

• Original Cundall documentation.
• Documentation for Aston code.

In our version, Trubal uses the output from LS algorithm as the initial configurations. Please refer to the following paper for details of this algorithm:

“A first-order phase transition defines the random close packing of hard spheres”, by Yuliang Jin and Hernan A. Makse, Physica A, Volume 389, 5362-5379 (2010), [pdf].

The LS part of the algorithm is used to generate packings with any volume fraction in the interval phi=0.55 (random loose packing) to phi=0.74, FCC. However, the LS algorithm does not generate mechanically stable packings as they are composed of hard spheres with no forces: A jammed packing in LS is an infinite kinetic pressure state (jamming is achieved with infinite number of collision per unit time). To find the forces that satisfy mechanical equilibrium, we run the MD part of the code applying an infinitesimal deformation to the particles in the LS packings following Hertz-Mindlin forces. Thus we generate packings in the limit of hard-sphere (zero stress pressure) satisfying force and torque balance in the entire range of volume fraction from RLP to FCC.

For more information about Trubal, please refer below to “Research on the phase diagram for jammed matter” and “Research on the entropy for jammed matter”.

Algorithm to the simulation of preparing the jammed packings and the calculation of the entropy from the packings. See these instructions for comments on how to use it.

• Download datasets we used in our research on the the entropy for different friction coefficient.
  • Dataset for friction coefficient 0.00001.
  • Dataset for friction coefficient 0.05.
  • Dataset for friction coefficient 0.1.
  • Dataset for friction coefficient 0.2.
  • Dataset for friction coefficient 0.5.
  • Dataset for friction coefficient 1.
  • Dataset for friction coefficient 10000.
  • Origin files for all the datasets.

Algorithm and instruction to find all the minima and transition states in clusters of small number of particles interacting via Lennard-Jones or Hertz potentials. The algorithms follows improved conjugate gradient methods to find all the minima, that is that finding the mechanical stable packings. The algorithm also find the value of the angoricity of the packings. This code has been used in the paper Wang, Song, Wang and Makse, Angoricity and compactivity describe the jamming transition in soft particulate matter, EPL 91, 68001 (2010).

One data example that we used in the paper to find the minima for packings with volume fraction 0.61.

This zip file contains all the experimental data on colloids and granular matter use to calculate the effective temperature of a colloidal glass and sheared granular matter from the particle trajectories. See the papers: P. Wang, C. Song, and H. A. Makse, Dynamic particle tracking reveals the aging temperature of a colloidal glass, Nature Physics 2, 526-531 (2006), [pdf] and C. Song, P. Wang, and H, A. Makse, Experimental measurement of an effective temperature for jammed granular materials, PNAS, 102, 2299 (2005), [pdf].

This tar file contains the program to generate long-range correlations in 1d and 2d. The method is explained in the paper: H. A. Makse, S. Havlin, M. Schwartz, and H. E. Stanley, Method for Generating Long-Range Correlations for Large Systems, Phys. Rev. E 53, 5445-5449 (1996).