Goals for 9/20

[ebridge2]

@ebridge2

• [x] begin adding documentation for ndmg-fngs leg. Deliverable: tbd. Documentation for single iter of pipeline and explanation of all functions. Documentation for how to use for running with multiple graphs. Documentation for how to use MNR plotting functions. Deliverable: these documents (1 week), and have 1 group member (2 weeks) go through starting with nifti files I give them and ultimately replicate the plots I produced. Deliverable: singlesub, multisub. Accidentally had a typo here in the hyperlink, which is why this is being updated right now. See the timestamps on the deliverables to confirm that this was all done/pushed at the appropriate time (before 5 PM today).
• [x] once simulations finish, send jovo plot results, and then begin drafting together a summary of what we've done that isn't 40 pages (figure reduction!) Deliverable: updated-model

bonus goals/stretch goals:

• [x] make changes for ndmg-fngs reflected in the issues in neurodata/ndmg for auto-merge.
• [x] get Cameron fMRI data before brain hacks. Upload our graphs to an s3 bucket for fcp-indi.
• [ ] make markdown file with fmri acquisition settings for CoRR datasets and email to shangsi.
• [ ] add information somewhere about fMRI algorithms used for discriminability paper. goal for next week, but here for safe keeping so I dont forget to mention it in meeting.

@ewalke31

• [ ] fix C observation matrix and ensure univariate EM KF algorithm works
• [ ] implement multivariate EM KF algorithm (normalize C in M-step) & provide examples
• [ ] investigate SVD and N4SID to properly initialize parameters & summarize
• [x] create R markdown to demonstrate above issues have been completed:

http://neurodatadesign.github.io/fngs/ewok_weekly/ewok_week_919.html (still having trouble with EM KF algorithm)

@ronnyb29

• [x] read n4sid and take notes
• [x] EM algorithm for system identification with Kalman filter. Implement and demonstrate that it works

Jupyter notebook with notes for both: https://github.com/Upward-Spiral-Science/fngs/blob/gh-pages/ron_weekly/sept19/EM_LDS.ipynb Had some trouble with implementing the EM for Kalman filtering, will seek help

@02agarwalt

• [x] email anthony about CIS account
• [ ] check out code https://github.com/Upward-Spiral-Science/fngs/blob/master/data_processing/Rutils/check_mnr.R here once you have ^ done, and then let me know and I will give you a short lesson about timeseries and how to use this analysis code via vchat maybe sometime this weekend if you have time
• [ ] go through as many lectures as you need of the kalman filter video series I slack'd to you to understand the basics of how it work, then try implementing it in either R (preferred) or python; I will cover in a vchat later this week a general framework for testing things that we try to adhere to

Stretch goals for everyone if time permits:

• [ ] Do first chapter of probability primer (https://www.youtube.com/playlist?list=PL17567A1A3F5DB5E4) and take notes
• [ ] Read Leo Breiman paper on random forest and take notes.

@jovo What probability textbook should we read?

[ronboger]

@gkiar Goals for the week

[gkiar]

What are the DoD's? I.e. what will you show me?

Also, per our walk after class, @ebridge2, you should also look into:

• doing LOL, PCA, for classification of covariance matrices
• making scree plots of covariance matrices
• refining your plots to be more concise and directed (i.e. what point does this illustrate for the reader?)

Greg Kiar gkiar07@gmail.com

On Mon, Sep 12, 2016 at 6:11 PM, ron boger notifications@github.com wrote:

@gkiar https://github.com/gkiar Goals for the week

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/Upward-Spiral-Science/fngs/issues/40#issuecomment-246511478, or mute the thread https://github.com/notifications/unsubscribe-auth/AEqDWNIExE80kZrU5Eusp5DMEKhaqEqrks5qpc4MgaJpZM4J6F6s .

[ronboger]

@gkiar if you're asking for me, i'll be showing you:

• notes on the n4sid paper
• notes on EM, code to implement it, and a graph of EM for system identification using Kalman filter if we get the implementation working

[gkiar]

👌