\n",
" 9/11 hijackers network\" src=\"pics/AviationNetwork.png\" width=\"770\">\n",
"
\n",
" Aviation network [Source: ICAO]\n",
"
SC 4125: Developing Data Products
\n", "Module-5: Network Science Toolkit
| \n",
" \n",
" | \n",
"
| \n",
" \n",
" | \n",
" \n",
" \n",
" | \n",
"
| \n",
" \n",
" | \n",
" \n",
" \n",
" | \n",
"
![\"Undirected](\"pics/graphtype-undirected.png\")
![\"Directed](\"pics/graphtype-directed.png\")
![\"Directed](\"pics/graphtype-multigraph.png\")
![\"Directed](\"pics/graphtype-multimode.png\")
![\"Directed](\"pics/GraphType-Bipartite.gif\")
![\"Sample](\"pics/top3-degreecentralnodes.png\")
![\"Sample](\"pics/top3-closeness-centrality.png\")
![\"Modularity\"](\"pics/modularity-graph.png\")
![\"Modularity\"](\"pics/FMeasureExample.png\")
\n",
"- See Evaluation of Clusterings -- Metrics and Visual Support by Achtert et al. for more discussions on comparing clustering results."
]
},
{
"cell_type": "markdown",
"metadata": {
"slideshow": {
"slide_type": "slide"
}
},
"source": [
"#### Ungraded Task 5.3\n",
"\n",
"> Compute the F1 score between the communities determined using modularity and label propagation methods.\n",
"> - Note that one of the limitations of F1 score is that it is unsuitable for fuzzy communities where same node may belong to multiple communities. "
]
},
{
"cell_type": "markdown",
"metadata": {
"slideshow": {
"slide_type": "slide"
}
},
"source": [
"## Link prediction\n",
"\n",
"> Predicting the existence/creation of a link between two entities in a network"
]
},
{
"cell_type": "markdown",
"metadata": {
"slideshow": {
"slide_type": "subslide"
}
},
"source": [
"#### Based on common neighbors\n",
"\n",
"- Underlying idea: Triadic closure\n",
" - Number of common neighbors\n",
" - Jaccard coefficient: Number of common neighbors normalized by total number of neighbors\n",
" * $\\frac{\\Gamma(u) \\cap \\Gamma(v)}{\\Gamma(u) \\cup \\Gamma(v)}$ where $\\Gamma(x)$ is the set of neighbors of node $x$ "
]
},
{
"cell_type": "code",
"execution_count": 13,
"metadata": {
"slideshow": {
"slide_type": "subslide"
}
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"[(2, 33, 6), (0, 33, 4), (7, 13, 4), (0, 32, 3), (1, 8, 3), (1, 33, 3), (2, 30, 3), (2, 31, 3), (4, 5, 3), (6, 10, 3), (8, 13, 3), (8, 31, 3), (13, 19, 3), (23, 31, 3), (27, 32, 3), (28, 32, 3), (0, 16, 2), (0, 28, 2), (0, 30, 2), (1, 12, 2)]\n"
]
}
],
"source": [
"# Common neighbors\n",
"# list node pairs which are not already neighbors\n",
"# ordered in decreasing number of common neighbors\n",
"topK=20\n",
"comm_neigh = [(e[0],e[1], len(list(nx.common_neighbors(G,e[0],e[1])))) for e in nx.non_edges(G)]\n",
"sortedCN = sorted(comm_neigh, key=lambda kv: kv[2],reverse=True)\n",
"print(sortedCN[0:topK])"
]
},
{
"cell_type": "code",
"execution_count": 14,
"metadata": {
"slideshow": {
"slide_type": "fragment"
}
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"[(14, 15, 1.0), (14, 18, 1.0), (14, 20, 1.0), (14, 22, 1.0), (15, 18, 1.0), (15, 20, 1.0), (15, 22, 1.0), (17, 21, 1.0), (18, 20, 1.0), (18, 22, 1.0), (20, 22, 1.0), (7, 13, 0.8), (4, 5, 0.75), (6, 10, 0.75), (9, 28, 0.6666666666666666), (17, 19, 0.6666666666666666), (19, 21, 0.6666666666666666), (13, 19, 0.6), (7, 12, 0.5), (7, 17, 0.5)]\n"
]
}
],
"source": [
"# Jaccard coefficient\n",
"jac_coeffs = list(nx.jaccard_coefficient(G))\n",
"# Notice that it transparently determines non-edges\n",
"sortedJC= sorted(jac_coeffs, key=lambda kv: kv[2],reverse=True)\n",
"print(sortedJC[0:topK])"
]
},
{
"cell_type": "code",
"execution_count": 15,
"metadata": {
"slideshow": {
"slide_type": "fragment"
}
},
"outputs": [
{
"data": {
"text/plain": [
"[(0, 1, 0.3888888888888889), (6, 16, 0.2)]"
]
},
"execution_count": 15,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"# If we want to compute the score for specific (set of) node pairs, that too can be done \n",
"# Example: nx.jaccard_coefficient(G, [(0, 1), (2, 3)])\n",
"list(nx.jaccard_coefficient(G, [(0, 1),(6, 16)]))\n",
"# Naturally, you can also compute this for node pairs which are adjacent, e.g., 6 and 16"
]
},
{
"cell_type": "markdown",
"metadata": {
"slideshow": {
"slide_type": "subslide"
}
},
"source": [
"#### Based on nodes' importances to their common neighbors individually\n",
"\n",
"- Underlying idea: How spread/concentrated are the ties \n",
" - Resource allocation\n",
" - Adamic-Adar"
]
},
{
"cell_type": "code",
"execution_count": 16,
"metadata": {
"slideshow": {
"slide_type": "fragment"
}
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"[(2, 33, 1.5666666666666664), (0, 33, 0.9), (1, 33, 0.7833333333333333), (4, 5, 0.6458333333333333), (6, 10, 0.6458333333333333), (23, 24, 0.5833333333333333), (25, 27, 0.5333333333333333), (0, 16, 0.5), (2, 31, 0.47916666666666663), (23, 31, 0.47549019607843135), (0, 32, 0.4666666666666667), (7, 13, 0.44027777777777777), (24, 33, 0.41666666666666663), (1, 8, 0.4125), (2, 30, 0.39444444444444443), (27, 31, 0.39215686274509803), (25, 32, 0.3666666666666667), (25, 33, 0.3666666666666667), (27, 32, 0.35882352941176476), (1, 32, 0.35)]\n"
]
}
],
"source": [
"# Resource Allocation\n",
"res_alloc = list(nx.resource_allocation_index(G))\n",
"sortedRA= sorted(res_alloc, key=lambda kv: kv[2],reverse=True)\n",
"print(sortedRA[0:topK])"
]
},
{
"cell_type": "code",
"execution_count": 17,
"metadata": {
"scrolled": true,
"slideshow": {
"slide_type": "fragment"
}
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"[(2, 33, 4.719381261461351), (0, 33, 2.7110197222973085), (1, 33, 2.252921681630931), (4, 5, 1.9922605072935597), (6, 10, 1.9922605072935597), (7, 13, 1.8081984819901584), (2, 31, 1.6733425912309228), (23, 31, 1.6656249548734432), (23, 24, 1.631586747071319), (0, 32, 1.613740043014111), (25, 27, 1.531574161186449), (1, 8, 1.5163157625699744), (2, 30, 1.4788841522548752), (0, 16, 1.4426950408889634), (27, 32, 1.4085855403276248), (28, 32, 1.34536123231926), (24, 33, 1.279458146995729), (27, 31, 1.2631953504915985), (25, 32, 1.179445561110859), (25, 33, 1.179445561110859)]\n"
]
}
],
"source": [
"# Adamic-Adar index\n",
"AdAd = list(nx.adamic_adar_index(G))\n",
"sortedAA= sorted(AdAd, key=lambda kv: kv[2],reverse=True)\n",
"print(sortedAA[0:topK])"
]
},
{
"cell_type": "markdown",
"metadata": {
"slideshow": {
"slide_type": "subslide"
}
},
"source": [
"#### Preferential attachment\n",
"- Underlying idea: More popular nodes are likely \"target of choice\" to connect with\n",
" * NetworkX implementation uses $|\\Gamma(u)|\\times|\\Gamma(v)|$ to determine preferential attachment score between nodes $u$ and $v$. \n",
" - This is essentially the product of their degree centralities\n",
" - Some other centrality scores too could be used instead \n"
]
},
{
"cell_type": "code",
"execution_count": 18,
"metadata": {
"slideshow": {
"slide_type": "fragment"
}
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"[(0, 33, 272), (0, 32, 192), (2, 33, 170), (1, 33, 153), (1, 32, 108), (3, 33, 102), (0, 23, 80), (3, 32, 72), (5, 33, 68), (6, 33, 68), (7, 33, 68), (0, 27, 64), (0, 29, 64), (0, 30, 64), (2, 31, 60), (13, 32, 60), (1, 31, 54), (4, 33, 51), (10, 33, 51), (24, 33, 51)]\n"
]
}
],
"source": [
"pref_at = nx.preferential_attachment(G)\n",
"sortedPA= sorted(pref_at, key=lambda kv: kv[2],reverse=True)\n",
"print(sortedPA[0:topK])"
]
},
{
"cell_type": "markdown",
"metadata": {
"slideshow": {
"slide_type": "subslide"
}
},
"source": [
"#### Using community information/side information\n",
"\n",
"- Underlying idea: Being part of a common community increases chance of linking up \n",
" - Soundarajan Hopcroft\n",
" * Variations: based on common neighbour count and resource allocation\n",
" - Within inter-cluster\n",
" * Compute the ratio of within- and inter-cluster common neighbors"
]
},
{
"cell_type": "code",
"execution_count": 19,
"metadata": {
"slideshow": {
"slide_type": "fragment"
}
},
"outputs": [],
"source": [
"# I am using the partitions created previosuly from Label Propagation algorithm\n",
"for node in G:\n",
" G.nodes[node][\"community\"] = colorsLP[node] "
]
},
{
"cell_type": "code",
"execution_count": 20,
"metadata": {
"scrolled": true,
"slideshow": {
"slide_type": "subslide"
}
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"[(2, 33, 11), (7, 13, 7), (27, 32, 6), (2, 30, 5), (13, 19, 5), (28, 32, 5), (0, 33, 4), (1, 12, 4), (2, 23, 4), (3, 17, 4), (3, 19, 4), (3, 21, 4), (7, 12, 4), (7, 17, 4), (7, 19, 4), (7, 21, 4), (8, 9, 4), (8, 14, 4), (8, 15, 4), (8, 18, 4)]\n",
"\n",
"[(2, 33, 1.3666666666666665), (27, 32, 0.35882352941176476), (7, 13, 0.3402777777777778), (2, 23, 0.3333333333333333), (23, 26, 0.3088235294117647), (26, 32, 0.3088235294117647), (2, 30, 0.2833333333333333), (27, 29, 0.25882352941176473), (1, 12, 0.22916666666666666), (7, 12, 0.22916666666666666), (12, 13, 0.22916666666666666), (3, 17, 0.1736111111111111), (3, 19, 0.1736111111111111), (3, 21, 0.1736111111111111), (7, 17, 0.1736111111111111), (7, 19, 0.1736111111111111), (7, 21, 0.1736111111111111), (13, 17, 0.1736111111111111), (13, 19, 0.1736111111111111), (13, 21, 0.1736111111111111)]\n"
]
}
],
"source": [
"SH_cn_coeff = nx.cn_soundarajan_hopcroft(G,community='community')\n",
"sortedSHcn=sorted(SH_cn_coeff, key=lambda kv: kv[2],reverse=True)\n",
"SH_ra_coeff = nx.ra_index_soundarajan_hopcroft(G,community='community')\n",
"sortedSHra=sorted(SH_ra_coeff, key=lambda kv: kv[2],reverse=True)\n",
"#preds.sort(key = operator.itemgetter(2),reverse=True)\n",
"print(sortedSHcn[0:topK])\n",
"print()\n",
"print(sortedSHra[0:topK])"
]
},
{
"cell_type": "code",
"execution_count": 21,
"metadata": {
"slideshow": {
"slide_type": "subslide"
}
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"[(27, 32, 3000.0), (1, 12, 2000.0), (2, 23, 2000.0), (3, 17, 2000.0), (3, 19, 2000.0), (3, 21, 2000.0), (7, 12, 2000.0), (7, 17, 2000.0), (7, 19, 2000.0), (7, 21, 2000.0), (8, 9, 2000.0), (8, 14, 2000.0), (8, 15, 2000.0), (8, 18, 2000.0), (8, 20, 2000.0), (8, 22, 2000.0), (8, 23, 2000.0), (8, 27, 2000.0), (8, 28, 2000.0), (8, 29, 2000.0)]\n"
]
}
],
"source": [
"IntClust_coeff=nx.within_inter_cluster(G, delta=0.001, community='community')\n",
"sortedIC=sorted(IntClust_coeff, key=lambda kv: kv[2],reverse=True)\n",
"print(sortedIC[0:topK]) "
]
},
{
"cell_type": "markdown",
"metadata": {
"slideshow": {
"slide_type": "slide"
}
},
"source": [
"#### Ungraded Task 5.4\n",
"\n",
"- Consider the HPD graph data provided in the HPD_full.net file\n",
" * Obtained from https://noesis.ikor.org/datasets/link-prediction\n",
"- Consider an simple version of the graph\n",
" * if needed, remove directionality, self-loops and multiple edges between any specific node pairs\n",
"- For the resulting graph, divide (randomly) ~90% edges to \"train\" your link predictor. Use ~10% edges to \"validate\" your predictions.\n",
" * You may want to use a seed for randomness, so that your experiments are reproducible\n",
"- Ungraded Task 5.4.1: Apply individually some of the basic link prediction approaches available in NetworkX, consider the top candidates (equal to the validation set size) to determine Recall and Precision of prediction per approach.\n",
"- Ungraded Task 5.4.2: Use the individual measures as features and apply machine learning technique(s) of your choice, to predict links. Report Recall, Precision (and any other metrics of interest) to determine the quality of your prediction. \n",
" * Make sure that you use the same 10% for comparison across subtasks.\n",
" * Though you may repeat the exercise with different subsets to study in/consistency. "
]
},
{
"cell_type": "code",
"execution_count": 22,
"metadata": {
"slideshow": {
"slide_type": "subslide"
}
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"8756 32331\n"
]
}
],
"source": [
"# To get you started: NetworkX has support to read several data formats \n",
"HPD_gr = nx.read_pajek(\"data/HPD_full.net\") # change path as suitable\n",
"HPD_ugr=nx.to_undirected(HPD_gr)\n",
"print(len(HPD_gr.nodes()),len(HPD_gr.edges()))"
]
},
{
"cell_type": "markdown",
"metadata": {
"slideshow": {
"slide_type": "slide"
}
},
"source": [
"# Neo4j\n",
"and Cypher\n",
"\n",
"\n",
"| \n",
" The following content on Neo4j and Cypher is (primarily) based on the book Graph Databases by Robinson et al. (specifically Chapter 3, where Cypher is discussed), and documentation resources on Neo4j website. Unless a different source is stated, the image snippets used in the following are from the stated book and Neo4j website. Neo4j provides for free this book as well as several other books on Graph DB, Algorithms and Data Science. Memgraph.com has a nice cheat sheet to get basic familiarity quickly with Cypher.\n", " \n", " \n", " \n", " | \n",
" \n",
" \n",
" | \n",
"
![\"Neo4j](\"pics/GraphDBs.png\")
\n",
"![\"Neo4j](\"pics/neo4jAuraPanel.png\")
"
]
},
{
"cell_type": "markdown",
"metadata": {
"slideshow": {
"slide_type": "slide"
}
},
"source": [
"### Data model\n",
"\n",
"- nodes\n",
" * properties: Nodes contain properties in the form of arbitrary key-value pairs.\n",
" * labels: Tags for the nodes, indicating role(s) the nodes play.\n",
"- relationships: A relationship in Neo4j is directed, with a start node and end node and a single name. \n",
" * properties: Similar to nodes, relationships can also have properties.\n",
" \n",
"![\"Simple](\"pics/neo4j-simplegraphpattern.png\")
"
]
},
{
"cell_type": "markdown",
"metadata": {
"slideshow": {
"slide_type": "subslide"
}
},
"source": [
"### Example\n",
" \n",
"\n",
"Flexibility\n",
"
| \n",
" \n",
" \n",
" | \n",
"
\n",
"\n",
"* specify some property values and node labels that help locate the relevant elements in the dataset\n",
"\n",
"> ![\"Simple](\"pics/neo4j-relationpattern.png\")
\n",
">\n",
"> ![\"Simple](\"pics/neo4j-bindingexample.png\")
\n",
" "
]
},
{
"cell_type": "markdown",
"metadata": {
"slideshow": {
"slide_type": "slide"
}
},
"source": [
"### Match and Return\n",
"\n",
"> ![\"Simple](\"pics/neo4j-matchexample1.png\")
"
]
},
{
"cell_type": "markdown",
"metadata": {
"slideshow": {
"slide_type": "subslide"
}
},
"source": [
"> \n",
"\n",
">* nodes are written inside parenthesis ()\n",
"> * node labels are prefixed by colons, e.g., :Person\n",
"> * the identifiers, e.g., a, b and c are on the left of the colon\n",
"> * use of identifier is optional: used if one needs to reference back to it \n",
"> * nodes can be anonymous in the query: e.g, (theater)<-[:VENUE]-() \n",
"> * relationships are written as --> OR <-- indicating directions\n",
"> * relationship names are indicated between the dashes, encapslated by square brackets, and prefixed by colon, e.g. -[:KNOWS]- \n",
"> * relationships or nodes may also be unnamed"
]
},
{
"cell_type": "markdown",
"metadata": {
"slideshow": {
"slide_type": "subslide"
}
},
"source": [
"> \n",
"\n",
">* curly braces are used to indicate node and relationship propery key-value pairs, e.g., {name:'Jim'}\n",
"> * in the example query, providing such a specific value creates an anchor in the graph, tying the query to a specific (set of) node instance(s)\n",
"> * underlying indexing helps optimize the search of the instances\n",
"> * one can explicitly ask to create indices also with ![\"Simple](\"pics/neo4j-matchexample2.png\")
\n",
"\n",
"> WHERE clause can more generally be used to describe constrains on match\n",
"> - e.g, WHERE p.born <= 1940"
]
},
{
"cell_type": "markdown",
"metadata": {
"slideshow": {
"slide_type": "subslide"
}
},
"source": [
"### Indeterminate paths\n",
"\n",
"> Queries with indeterminate relationships - both in terms of types, and distance within the graph can be composed. \n",
">\n",
"> ![\"Simple](\"pics/neo4j-matchexample3.png\")
"
]
},
{
"cell_type": "markdown",
"metadata": {
"slideshow": {
"slide_type": "slide"
}
},
"source": [
"#### Check the book & manual for numerous other Cypher options and features\n",
"\n",
"- Post-processing results: aggregation, ordering, filtering, etc.\n",
"- Query chaining (writing complex queries in simplers sub-parts)\n",
"- Common data modelling pitfalls\n",
"- Loading and modifying data\n",
" * Cypher instructions, e.g. CREATE\n",
" * Bulk data uplaod\n",
" * Import .csv files\n",
" * neo4j-admin import tool\n",
"- Database management, access control, etc. "
]
},
{
"cell_type": "markdown",
"metadata": {
"slideshow": {
"slide_type": "slide"
}
},
"source": [
"### Hands-on with Neo4j and Cypher\n",
"\n",
">I use Neo4j's cloud service Aura along with its default Movie data set for the demo codes that follow. \n",
"\n",
"> The example code snippets used here are based on and expands over Neo4j's introduction to Cypher and a nice and short Medium article on how to integrate Neo4j with Jupyter notebook, which use the same movie dataset. "
]
},
{
"cell_type": "code",
"execution_count": 23,
"metadata": {
"slideshow": {
"slide_type": "fragment"
}
},
"outputs": [],
"source": [
"#!pip install neo4j\n",
"#!pip install neo4j-driver==1.6.2\n",
"#!pip install --upgrade py2neo==2021.1.5 \n",
"## The latter version (2021.2.0) of py2neo unfortunately triggers errors with rest of my environment setup\n",
"## To use this code with rest of your config, you may have to play around with the neo4j driver and py2neo versions \n",
"# You may also need to restrat kernel after installations\n",
"import neo4j\n",
"from py2neo import Graph,Node,Relationship,NodeMatcher \n",
"# Also checkout https://pypi.org/project/ipython-cypher/ for using Cypher with Jupyter magic\n",
"## I have worked with it with local DB instance, however I could not make it work with Aura and my current configurations"
]
},
{
"cell_type": "code",
"execution_count": 24,
"metadata": {
"slideshow": {
"slide_type": "subslide"
}
},
"outputs": [],
"source": [
"# DB credentials obfuscated: Replace this portion with your own credentials\n",
"# Comment the following out, and populate your own \n",
"fo = open(\"neo4j-Aura-creds.txt\", \"r\")\n",
"strng = fo.readlines()\n",
"fo.close()\n",
"\n",
"URI=strng[0].strip()\n",
"UserID=strng[1].strip()\n",
"NeoPass=strng[2].strip()\n",
"\n",
"# Uncomment and populate your info based on your own instance, replacing the XXX\n",
"# URI=\"XXX\"\n",
"# UserID=\"XXX\"\n",
"# NeoPass=\"XXX\""
]
},
{
"cell_type": "code",
"execution_count": 25,
"metadata": {
"slideshow": {
"slide_type": "fragment"
}
},
"outputs": [],
"source": [
"# Connect to the Neo4j server \n",
"# URI could be as from the Aura cloud service, or your local instance\n",
"# For exposition, I am using Aura, along with the default \"Movie database\" \n",
"# My URI looks like: neo4j+s://XXX.databases.neo4j.io\n",
"\n",
"graph = Graph(URI, auth=(UserID, NeoPass))"
]
},
{
"cell_type": "markdown",
"metadata": {
"slideshow": {
"slide_type": "slide"
}
},
"source": [
"### The Movie dataset\n",
"\n",
"Default dataset with Aura"
]
},
{
"cell_type": "code",
"execution_count": 26,
"metadata": {
"slideshow": {
"slide_type": "fragment"
}
},
"outputs": [
{
"data": {
"text/html": [
"| \n", " | NodeType | \n", "
|---|---|
| 0 | \n", "[Movie] | \n", "
| 1 | \n", "[Person] | \n", "
| \n", " | NodeType | \n", "Properties | \n", "
|---|---|---|
| 0 | \n", "[Movie] | \n", "[tagline, title, released] | \n", "
| 1 | \n", "[Movie] | \n", "[title, released, tagline] | \n", "
| 2 | \n", "[Movie] | \n", "[released, tagline, title] | \n", "
| 3 | \n", "[Movie] | \n", "[title, released] | \n", "
| 4 | \n", "[Person] | \n", "[name, born] | \n", "
| 5 | \n", "[Person] | \n", "[name] | \n", "
| \n", " | RelationType | \n", "Properties | \n", "
|---|---|---|
| 0 | \n", "ACTED_IN | \n", "[roles] | \n", "
| 1 | \n", "PRODUCED | \n", "[] | \n", "
| 2 | \n", "DIRECTED | \n", "[] | \n", "
| 3 | \n", "WROTE | \n", "[] | \n", "
| 4 | \n", "REVIEWED | \n", "[summary, rating] | \n", "
| 5 | \n", "REVIEWED | \n", "[rating, summary] | \n", "
| 6 | \n", "FOLLOWS | \n", "[] | \n", "
| \n", " | PersonName | \n", "MovieName | \n", "
|---|---|---|
| 0 | \n", "Aaron Sorkin | \n", "A Few Good Men | \n", "
| 1 | \n", "Al Pacino | \n", "The Devil's Advocate | \n", "
| 2 | \n", "Annabella Sciorra | \n", "What Dreams May Come | \n", "
| 3 | \n", "Anthony Edwards | \n", "Top Gun | \n", "
| 4 | \n", "Audrey Tautou | \n", "The Da Vinci Code | \n", "
| ... | \n", "... | \n", "... | \n", "
| 167 | \n", "Victor Garber | \n", "Sleepless in Seattle | \n", "
| 168 | \n", "Werner Herzog | \n", "What Dreams May Come | \n", "
| 169 | \n", "Wil Wheaton | \n", "Stand By Me | \n", "
| 170 | \n", "Zach Grenier | \n", "RescueDawn | \n", "
| 171 | \n", "Zach Grenier | \n", "Twister | \n", "
172 rows × 2 columns
\n", "| \n", " | PersonName | \n", "Colleague | \n", "
|---|---|---|
| 0 | \n", "Kevin Bacon | \n", "Aaron Sorkin | \n", "
| 1 | \n", "Kevin Bacon | \n", "Bill Paxton | \n", "
| 2 | \n", "Kevin Bacon | \n", "Christopher Guest | \n", "
| 3 | \n", "Kevin Bacon | \n", "Cuba Gooding Jr. | \n", "
| 4 | \n", "Kevin Bacon | \n", "Demi Moore | \n", "
| 5 | \n", "Kevin Bacon | \n", "Ed Harris | \n", "
| 6 | \n", "Kevin Bacon | \n", "Frank Langella | \n", "
| 7 | \n", "Kevin Bacon | \n", "Gary Sinise | \n", "
| 8 | \n", "Kevin Bacon | \n", "J.T. Walsh | \n", "
| 9 | \n", "Kevin Bacon | \n", "Jack Nicholson | \n", "
| 10 | \n", "Kevin Bacon | \n", "James Marshall | \n", "
| 11 | \n", "Kevin Bacon | \n", "Kevin Pollak | \n", "
| 12 | \n", "Kevin Bacon | \n", "Kiefer Sutherland | \n", "
| 13 | \n", "Kevin Bacon | \n", "Michael Sheen | \n", "
| 14 | \n", "Kevin Bacon | \n", "Noah Wyle | \n", "
| 15 | \n", "Kevin Bacon | \n", "Oliver Platt | \n", "
| 16 | \n", "Kevin Bacon | \n", "Rob Reiner | \n", "
| 17 | \n", "Kevin Bacon | \n", "Ron Howard | \n", "
| 18 | \n", "Kevin Bacon | \n", "Sam Rockwell | \n", "
| 19 | \n", "Kevin Bacon | \n", "Tom Cruise | \n", "
| 20 | \n", "Kevin Bacon | \n", "Tom Hanks | \n", "
| \n", " | PersonName | \n", "CoActor | \n", "
|---|---|---|
| 0 | \n", "Kevin Bacon | \n", "Aaron Sorkin | \n", "
| 1 | \n", "Kevin Bacon | \n", "Bill Paxton | \n", "
| 2 | \n", "Kevin Bacon | \n", "Christopher Guest | \n", "
| 3 | \n", "Kevin Bacon | \n", "Cuba Gooding Jr. | \n", "
| 4 | \n", "Kevin Bacon | \n", "Demi Moore | \n", "
| 5 | \n", "Kevin Bacon | \n", "Ed Harris | \n", "
| 6 | \n", "Kevin Bacon | \n", "Frank Langella | \n", "
| 7 | \n", "Kevin Bacon | \n", "Gary Sinise | \n", "
| 8 | \n", "Kevin Bacon | \n", "J.T. Walsh | \n", "
| 9 | \n", "Kevin Bacon | \n", "Jack Nicholson | \n", "
| 10 | \n", "Kevin Bacon | \n", "James Marshall | \n", "
| 11 | \n", "Kevin Bacon | \n", "Kevin Pollak | \n", "
| 12 | \n", "Kevin Bacon | \n", "Kiefer Sutherland | \n", "
| 13 | \n", "Kevin Bacon | \n", "Michael Sheen | \n", "
| 14 | \n", "Kevin Bacon | \n", "Noah Wyle | \n", "
| 15 | \n", "Kevin Bacon | \n", "Oliver Platt | \n", "
| 16 | \n", "Kevin Bacon | \n", "Sam Rockwell | \n", "
| 17 | \n", "Kevin Bacon | \n", "Tom Cruise | \n", "
| 18 | \n", "Kevin Bacon | \n", "Tom Hanks | \n", "
| \n", " | PersonName | \n", "Colleague | \n", "
|---|---|---|
| 0 | \n", "Kevin Bacon | \n", "Rob Reiner | \n", "
| 1 | \n", "Kevin Bacon | \n", "Ron Howard | \n", "
| \n", " | PersonName | \n", "Colleague | \n", "
|---|---|---|
| 0 | \n", "Kevin Bacon | \n", "Rob Reiner | \n", "
| 1 | \n", "Kevin Bacon | \n", "Ron Howard | \n", "
| \n", " | tagline | \n", "title | \n", "released | \n", "
|---|---|---|---|
| 0 | \n", "In every life there comes a time when that thi... | \n", "That Thing You Do | \n", "1996 | \n", "
| 1 | \n", "Come as you are | \n", "The Birdcage | \n", "1996 | \n", "
| 2 | \n", "Break The Codes | \n", "The Da Vinci Code | \n", "2006 | \n", "
| 3 | \n", "Evil has its winning ways | \n", "The Devil's Advocate | \n", "1997 | \n", "
| 4 | \n", "Walk a mile you'll never forget. | \n", "The Green Mile | \n", "1999 | \n", "
| 5 | \n", "Welcome to the Real World | \n", "The Matrix | \n", "1999 | \n", "
| 6 | \n", "Free your mind | \n", "The Matrix Reloaded | \n", "2003 | \n", "
| 7 | \n", "Everything that has a beginning has an end | \n", "The Matrix Revolutions | \n", "2003 | \n", "
| 8 | \n", "This Holiday Season… Believe | \n", "The Polar Express | \n", "2004 | \n", "
| 9 | \n", "Pain heals, Chicks dig scars... Glory lasts fo... | \n", "The Replacements | \n", "2000 | \n", "
| 10 | \n", "I feel the need, the need for speed. | \n", "Top Gun | \n", "1986 | \n", "
| 11 | \n", "Don't Breathe. Don't Look Back. | \n", "Twister | \n", "1996 | \n", "
![\"Flexible:](\"pics/tree-cycle.png\")
\n",
"![\"Arms](\"pics/armsnetwork.png\")
\n",
" ![\"US](\"pics/US-Facebook-AudienceGroups.png\")
\n",
" ![\"Bitcoin](\"pics/BitcoinFlow.png\")
\n",
" ![\"Neo4j](\"pics/neo4j-bookcover.png\")
\n",
" ![\"Flexible:](\"pics/neo4j-crossdomain.png\")
\n",
"