sagarjauhari / graph_anomaly_detect Goto Github PK

Your implementation should write the output to a file using the following format:
• The first line should indicate the upper threshold value and the lower threshold value, separated by a space. If your algorithm only requires one threshold (lower or upper), then use NaN as the value of the non-required threshold.
• The rest of the lines should indicate each anomalous time point and its corresponding value in the time series, separated by a space. If your algorithm calculates two values per time point
For example:

upper threshold [space] lower thresohld
time point i [space] value of time point i and i-1 [space] value of time point i and i+1
time point j [space] value of time point j and j-1 [space] value of time point j and j+1

Will display once project is opensourced.

Description of readme file has been given in the project description:

README file with detailed instruction. It should contain at least the following:
(a) Software that needs to be installed (if any) with URL’s to download it from and instructions on how to install them.
(b) Environment variable settings (if any) and OS it should/could run on.
(c) Instructions on how to run the program (and the script to change the output format, if included).
(d) Instructions on how to interpret the results.
(e) Sample input and output files.
(f) Citations to any software you may have used or any dataset you may have tested your ode on.

No need for creating new branch. Just assign yourself to the task and edit as is.

induced_subgraph seems to be too intensive

For 'reality_mining_voices':

Aggregating features
/usr/lib/python2.7/dist-packages/scipy/stats/stats.py:1067: RuntimeWarning: invalid value encountered in double_scalars
  vals = np.where(zero, 0, m3 / m2**1.5)

This function is taking 43% of the total time

Current implementation is taking about 4 hours for graph p2p-Gnutella!