06 November 2013, Meeting

[lauraccunningham]

1200, Kresge

[lauraccunningham]

Questions:

How can we cut down on computational time in R? The SAPP & PtProcess packages take an exorbitant amount of time to run the algorithms, and we need to speed this up. We've considered "unrolling" and vectorizing, but are unsure if Python will help speed up this process further! Can you give us any insight on how we can fix this with the given time constraints?

@HeXXiiiZ @aculich

[lauraccunningham]

Review

Timeline for Class

• November
  • 07, November
    • Pick a Domain to Standard Testing of Data, Preliminarily
  • 12, Tuesday
    • Check-In 1: Preliminary Analysis
      • What has been accomplished so far?
      • What needs to be completed by Check-In 2?
      • What needs to be completed by Check-In 3?
  • 14, Thursday
    • Check-In 1.5: Preliminary Analysis
  • 19, Tuesday
    • Check-In 2: Modeling
      • What has been accomplished so far?
      • What needs to be completed by Check-In 3?
  • 21, Thursday
  • 26, Tuesday
    • Check-In 3: Diagnostics & Validity
      • Are you completed with your portion of the final project?
        • If not, what is needed to complete sub-group's portion of the project? Any groups unfinished at this point will need to completed entirely by Monday, December 2, 23:59:59.
  • 28, Thursday
    • Thanksgiving, No Class
• December
  • 03, Tuesday
    • All Subgroup's will need to have completed their portions of the assignment by midnight the night before
      • Required Completed Material:
        • Gather all coding materials from Horizontal Groups within a single iPython Notebook
        • Gather all instructions for Reproducibility from Presenters
          • Include any write-ups and information gathered from sub-groupsto better understand the models and outcome
  • 05, Thursday
    • Compile all Materials Required to Complete Presentation
    • Compile Paper of Project: Process, Actions, & Outcome
    • Presenters - Compile Presentation & Designate all roles for Presentation & Expectations of Presenters (A meeting with Aaron will be necessary to prepare for the final presentation & table at the Data Science Faire.)
  • 10, Tuesday, Dead Week
    • Presentation Run-Through
      • To Run-through Presentation for the 1st Annual Data Science Faire hosted by the Berkeley Institute for Data Science (BIDS)
      • Evaluation of Presentation by Peers as Audience, Peer Review & Critique (Attendance of all class members will be needed to be successful and necessary.)
  • 12, Thursday, Dead Week
    • Data Science Faire, 1st Annual, Presentation: TIME & LOCATION TBD

Analyzer's Timeline Based on Needs & Goals

• November
  • 07, Thursday
    • Pick a Domain: Data Set for all groups to being looking at either a state, region, fault line, etc.
    • Communicate With Other Analyzers to Work in Parallel
  • 12, Tuesday
    • Check-In 1:
      • Understand Packages in R
      • Begin analysis of Curator Data with SAPP and PtProcess
        • Vectorizing, Make Improvements upon Algorithms / Computation
        • Writing into Python to Speed Up Algorithm (If Time Allows - Consult with Chris)
  • 14, Thursday
    • Check-In 1.5:
      • Fitting the Models with Preliminary Analysis
      • Comparisons of models: MDA v. ETAS
      • (Simlutating Earthquake Catalogs, Permute Catalogs - Not Necessary)
  • 19, Tuesday
    • Check-In 2
      • Generating Alarms
      • Find Prediction Accuracy per tau
      • All ETAS complete to begin looking at MDA
        • Generate & Examine Error diagrams
        • Provided Error Rates and Data Points of Outputs will be prepared for Visualizers
  • 21, Thursday
  • 26, Tuesday
    • Check-In 3
      • Iterate MDA Model to beat ETAS with a More Accurate Earthquake Alarm Error Model
  • 28, Thursday
    • Thanksgiving, No Class
• December
  • 03, Tuesday
    • Send All Information Gathered and Write-Up to Presenters
  • 05, Thursday
  • 10, Tuesday, Dead Week
    • Attend Presentation Run-Through
      • Provide Feedback and Constructive Criticism
  • 12, Thursday, Dead Week
    • Data Science Faire, 1st Annual, Presentation: TIME & LOCATION TBD

Things to Beginning Looking at, The Big Picture
Journal Format

• Preliminary Analysis
• Modeling
• Diagnostic
• Validity