On Tuesday, I had a team meeting with Roger and Arun. In the meeting, Roger shared the basic concept of causality and the overview of HPCC system causality toolkit.
Other than that, in these two days, I mainly worked on the test examples in Because module.
Synth submodule: By running genTest.py and viewing the code in gen_data.py, understand the process of generating synthesis data, and how to write the model description to utilize this power tool. It is the basic for the experiments next.
Probability submodule: By running test examples, understand the capability of main class ProbSpace, such as the calculation of the probability or conditional probability for any combinations of variables, approximation of the distribution of variables, calculating basic statistics of distribution, and the plot of PDF. Details are still need to be examined, like the details of DP, JP, UP to calculate conditional PDF, how to calculate dependency, the effect of Power parameter.
Causality submodule: By running test examples, understand the capability of main class cGraph, which combines the hypothesized causal model and data. PropSpace is applied on data to have statistic calculation. Graph algorithm is applied on causal model to calculate causal related formula like do operation, and deduce independence and dependence. It can calculate ACE, CDE and CIE, scan data to find causal relationship, and validate the hypothesized causal model by independence and dependence relationships between variables. Implementation details still need to be check, like how to do the scan on graph, how to deduce dependence relationship from graph.
Some questions generated when doing the test and will be discussed further, some typo are discovered in code and will be amended.
No comments:
Post a Comment