Difference between revisions of "Complex Systems Summer School 2016-Tutorials"
From Santa Fe Institute Events Wiki
(→Structural pattern discovery in spatio-temporal data)
(→Introduction to Git)
|Line 16:||Line 16:|
Revision as of 23:51, 14 June 2016
Date & Time:
Motivation and content:
Introduction to Git
Date & Time: To be convened with interested participants
Motivation and content: An introduction to management and versioning of source code using the basic commands of the Git system. As your model grows, the source code becomes increasingly more complex and difficult to maintain. Git takes care of this workload by recording every change to your project files, helping you to try and experiment new model features in a safe way, tracking your model variants and versions, and implementing collaboration with your co-researchers. This tutorial will cover the basics of: commits, branches, merging, versioning, and collaboration.
Suggestions: You should have basic experience using text terminals. Please bring your own laptop. You are invited to open a free account on gitlab.com before coming to the tutorial. Please contact email@example.com with any inquiries.
Structural pattern discovery in spatio-temporal data
Speaker: [Pavel Senin]
Date & Time: TBD
Motivation and content: The problem of informative pattern discovery in time series traditionally receives much attention. Discovering patterns is important in areas as diverse as medicine, security, astronomy, industry, sciences, and finance, to name just a few, where patterns typically convey critical and actionable information.
In this hands-on tutorial, I review three symbolic discretization-based techniques for time-series patterns discovery: (i) SAX-VSM -- an algorithm for discovery of class-characteristic patterns in contrast time series sets, (ii) GrammarViz -- a grammatical inference-based technique for variable-length time series motif (i.e., frequent pattern) discovery, and (iii) RRA -- a grammatical inference and algorithmic (i.e., Kolmogorov) complexity -based algorithm for variable-length time series discord (i.e., anomaly) discovery. In addition, I discuss our recent effort for discretization parameters optimization.
All discussed techniques are implemented in Java (high-throughput) and R (convenience) and I show how to use these in experimental/exploratory settings.
1) SAX-VSM: Interpretable Time Series Classification Using SAX and Vector Space Model
2) GrammarViz 2.0: a tool for grammar-based pattern discovery in time series
3) Time series anomaly discovery with grammar-based compression