ernaneluis / graph-dynamics

graph-dynamics lib

graph-dynamics

graph-dynamics lib

graph-dynamics's Issues

Create MacroStates Run Parameters File

Small empirical network

Please create the network of computer science or something small so we are able to make experiments quickly in something relatively small

Bipartite Transaction Graphs

Once the MongoDBHandlers is migrated please include in the aggregation

/MongoDBHandlers/AggreagateMongoDB.ipynb

Below the aggregation markup the identifier of the transaction, since that will allow us to create bipartegraph if necessary.

Temporal Patterns of Our Model

Modify the TX model so that the activity changes every time, keep the number of neighbors constant. Preferably assign the activity as a function the amount of money.

Data

• data from jan to july 2017 by day (INCOMPLETE MUST REDO)

---

Create a model step

• create the activity function where we can play with that dependents money time and some type of guy

---

Analysis step

• run the motif of the real data and compare with simulation motif
• check with the motif use the same transaction network
  From the paper:

BITCOIN. Bitcoin is a decentralized digital currency and payment
system. This dataset consists of all payments made up to October
19, 2014 [11]. Nodes in the network correspond to Bitcoin addresses,
and an individual may have several addresses. An edge
(u, v, t) signifies that bitcoin was transferred from address u to address
v at time t.
The gains of the fast
algorithm are the largest for BITCOIN, which is due to some pairs
of nodes having many edges between them and also participating
in many triangles
Cycles in BITCOIN. Of the eight 3-edge triangle motifs, M2,4 and
M3,5 are cyclic, i.e., the target of each edge serves as the source
of another edge. We observe in Fig. 8 that the fraction of triangles
that are cyclic is much higher in BITCOIN compared to any
other dataset. This can be attributed to the transactional nature of
BITCOIN where the total amount of bitcoin is limited. Since remittance
(outgoing edges) is typically associated wit

edges constructed from the transactions: a transaction with 2 inputs and 3 outputs results in 6 edges (all possible combinations), an edge may appear multiple times, with the corresponding transaction IDs http://www.vo.elte.hu/bitcoin/downloads.htm#database

• macro state of wealth distribution of nodes by time
• how to find out how the omega w, of big clam changes over time by doing histogram by time

---

traninig step

• bigclam by one community where the initial community is every node

---

prediction step

• [ ]

---

let there be trade inferring the dynamics of the time-varying bitcoin transaction network

TxGraph and Perra Graph be subclass of Graph class

Hierarchical

Please check in the drive the papers relating hierarchies and centralities.

Create graphs using the code:

https://github.com/ernaneluis/graph-dynamics/blob/master/Notebooks/Networks/Communities/HierarchicalGraphs.ipynb

And try to see if the nodes embeddings recover the different levels of the hierarchy. If not, try to modify the random walks until they do

Time Series BitCoin transactions

Time Series of:

• Nodes
• Edges
• new nodes
• per node weights of big clam (activity)
• number of triangles and clustering coefficient

Apply The Temporal Network Motif Algorithms

Time delta as part of parameters file

Create Tensor for the Multiplex Representation of the Dynamics

Just follow

https://drive.google.com/open?id=0B5P9ajMXC2Y0THk2STVQS3FIdnM

Change print to loggers

JSON parameter for Finite Process Constructor

Migrate the graph analysis codes to graph dynamics

No need to use the abstract classes for now

Perform the multilabel Classification

For start use:

https://snap.stanford.edu/data/cit-HepTh.html

During the training phase, we observe a certain fraction of the nodes and all their labels.
The task is to predict the labels for the remaining nodes.

The node features representations are input to a one-vs-rest logist regression classifier with L2 regularization.

Finite Beta Process

The theory regarding the finite process is located in:

https://drive.google.com/open?id=0B5P9ajMXC2Y0ZFV1QlYyNU51Z1E

The simple beta process is just an example that serve as a motivation for the thing. In principle you should create something similar to:

https://github.com/ernaneluis/graph-dynamics/blob/master/graph_dynamics/random_measures/finite_process.py#L16

JSON parameter for TxGraph constructor

Create website

https://pages.github.com/

Quantidy Centrality Evolution Change

Fit exponential and polynomials to the centrality evolutions.

In order to evaluate the fit use the function defined in the following paper:

https://drive.google.com/open?id=0B5P9ajMXC2Y0eDdFYmtxZXdJVHc

There you can see the variance which is simply the error penalized with the number of parameters

Finish Palla Dynamics

Evolution must incorporate the N variables corresponding to the time dependance of the edges:

https://github.com/ernaneluis/graph-dynamics/blob/master/graph_dynamics/dynamics/PittWalker.py#L87

Centrality Plot

recreate the plots shown:

https://en.wikipedia.org/wiki/Centrality

Binary Matrix

Obtain binary matrix from Beta process:

Take a look at:

https://drive.google.com/open?id=0B1WmdewksxI9WUE0X01feU9OQ2s

Layered Plot

Create a 3D plot for the graph time series, where the Y dimension is time and X and Z correspond to the position of the nodes.

Similar to https://matplotlib.org/examples/mplot3d/bars3d_demo.html

only that instead of a bar plot, you have a graph from the graph time series

Profiling

Please locate the slowest functions or methods in

https://github.com/ernaneluis/graph-dynamics/blob/master/tests/dynamicsTest.py#L21

Create a Temporal Representation for The TxDynamics

one graph in which the edges have attributes
when they where created (and if possible when they where destroyed)
that way instead of a list of graph
we have just one graph

https://github.com/ernaneluis/graph-dynamics/blob/master/graph_dynamics/dynamics/GraphPathsHandlers.py#L19

Perform the Link prediction

In Link prediction, we are given a network with a certain fraction of edges removed,
and we would like to predict these missing edges. We generate the labeled data set as follows:
To obtain positive examples, we remove 50% of the edges chosen randomly from the network while ensuring that the residual network obtained after the edge removals is connected,

DO THE PREDICTION FOR THE FOLLOWING DATA SET:

https://snap.stanford.edu/data/cit-HepTh.html

3D printing

be a modular function of dynamics

Plot Test and Training Set for Link Prediction

BitCoin Graph

Please define a class for the bitcoin models, we need t simply to finish this class:
https://github.com/ernaneluis/graph-dynamics/blob/master/graph_dynamics/networks/generative_graphs.py#L16

Called simply bitcoin graph or ernane graph since is not goin to be generate but to define a graph instead

be need a framework like

https://github.com/ernaneluis/graph-dynamics/blob/master/tests/dynamicsTest.py#L35

So we can define a graph and then evolve it. This will require an abstract class called Graphs or something like that, but I will write it down after you work on the particular example

generate graphs path window

This function should have as parameters the time window required for the evolution.

Train the Big Clam

Train the big clam per (community number ) snapshot for the real transaction network, this should be a function applied to dynamics.

use the bigclam to find out the activz rate to be the input parameter of the perra graph

and return a times series of the bigclam edges(activity rate) matrix to be a dynamic function

Create Dynamical Class For Perra Dynamics

We need to define a dynamical class and graph for the simple Parra Model

https://drive.google.com/open?id=0B5P9ajMXC2Y0NTd6QkhMOGFvWDQ

This should be a simple version of the already existing Tx Dynamics

CReate Thesis Repo for Latex and Writting

• issues to be created there:
  1. Explain the bit transaction network
  2. Explain the Perra dynamics
  3. Explain the Bit Clam

https://www.amazon.com/Money-Unauthorized-Biography-Coinage-Cryptocurrencies/dp/0345803558/ref=sr_1_1?ie=UTF8&qid=1499779663&sr=8-1&keywords=money+the+unauthorized+biography

Create Temporal Graph from Series

This will require a detailed series of edges from the graph series in order to perform a posterior transformation or to include this directly in the dynamics