week 04/24

[ewalke31]

@ewalke31

• [x] for Thursday: continue w/ progress on documentation (make links at top in table & add relevant schematic section) https://neurodatadesign.github.io/fngs/ewalke31/tutorials/week_0410/Schematic.html
• [x] for Monday: finish documentation for the second half of the pipeline https://neurodatadesign.github.io/fngs/ewalke31/tutorials/week_0410/Schematic.html

@ebridge2

• [x] for Thursday: connectomes as thresholding top 60% edges, but retain original edge-weights. Reanalyze and produce qa on 2 datasets (BNU1 and kara's data). BNU1 aal BNU1 desikan kara desikan note: relative probabilities are not correct. this is because number of nonzero edges given that we explicitly threshold away to make all have the same number of nonzero edges makes more vector rank deficient, and therefore we cannot compute a kde as it would have infinite weight at the particular threshold we chose. to fix this, I just input a blank value of [0, 1].
• [x] pivot: update registration instead of below 2 deliverables. This pivot was approved by Jovo. Deliverable: report performance of registration after breaking apart the timeseries. Quantitative comparison of pre-split registration, and post-split registration, to see how performance is impacted. specs
• [ ] stretch: get EC2 working.

moved to next week:

• [ ] for Monday: batch calls possible from website.
• [x] for Monday: update pipeline schematic with figures from the qa produced by FNGS. schematic

non-class goal:

• [ ] update presentation.

@02agarwalt

• [x] finalize AWS batch implementation for pipeline: set local environment variables with credentials check user's batch and create job definition, queue, etc. fix batch deployment (hard drive specs, etc.) alert user with submitted job ids Link Problems: https://forums.aws.amazon.com/thread.jspa?messageID=780181 https://forums.aws.amazon.com/thread.jspa?messageID=771333

[jovo]

@ebridge2 @02agarwalt anything?

[02agarwalt]

updated

[ebridge2]

thoughts for kara stuff: get subject-specific covariates from kara. make inferences we are making for each task across covariates like sex once we can figure out how to best show the model works.

[jovo]

let's focus on the prize. we have data and question, let's get method that answers it...

On Thu, Apr 20, 2017 at 1:08 AM, Eric Bridgeford notifications@github.com wrote:

thoughts for kara stuff: get subject-specific covariates from kara. make inferences we are making for each task across covariates like sex once we can figure out how to best show the model works.

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/NeuroDataDesign/fngs/issues/13#issuecomment-295582498, or mute the thread https://github.com/notifications/unsubscribe-auth/AACjcneY8qrmisDYTL_u7MlcnL77DcW_ks5rxug1gaJpZM4M_o6u .

-- the glass is all full: half water, half air. neurodata.io ps - i tend to respond to email weekly so i can spend more time with my wife & baby, thank you

[jovo]

@jovo feedback

@ebridge2 split up epi_reg to keep under 30 GB