# grant weekly meeting notes
###### tags: `grant`

## First meeting: 3/10/22

Workplan 1: maintenance and docs

Workplan 2: algorithmic functionality

Workplan 3: neuro and vitals

Start plan: look for easy issues and PRs to get in.

review
1. https://github.com/sktime/sktime/pull/3021

address

2. https://github.com/sktime/sktime/issues/2132
3. https://github.com/sktime/sktime/issues/2116

Use Trello: https://trello.com/b/tkfFXiWJ/sktime

## 10/10/22

1. Review and discuss PRs
2. Review other to dos
3. Plan next steps

1. Review PRs

#3431 catch22, related to whole nested_univ issue, low priority, note to chase for re-review
#3546 load from tsc.com. blocked on testing, franz chased
#3523 notebook scikit learn examples, chased 5/10, chase again 19/10
#3528 add testing for contractable and train_estimate. Blocked because of the design of get_train_estimates, chase 17/10.
#3529 kernel -> convolution, franz wants a strange wrapper approach, leave it.
#3532 write to ts, blocked because of config, requested remove comment on 6/10, rerequest 20/10.
#3546: todo Tony: review docs for resampling

2. Other to dos

trello lists: set up board for each workpage
debrief on turing and UCL.

bake off redux: ongoing. Tony with MTSC, matt doing UTSC.

3. Next step

two medium term goals:
1. sort out port of sktime-dl
2. work out the read the docs works, how to debug/visualise changes

todo: work out the read the docs works, how to debug/visualise changes (Matt)
todo: Run est-eval using sktime fork main (Matt)

todo: Tony to set off a single job on kraken with 1 njobs and again with 90 odd

## 17/10/22

### Open PRs
Review of https://trello.com/b/YCi6gsoc/sktime-pull-requests
#3021 load from UEA, ake to next level on 24/10/22 if still blocked
#3431 catch22, take to next level on 24/10/22 if still blocked
#3523 notebook scikit learn examples, next level on 24/10/22
#3528 add testing for contractable and train_estimate. Seems to be failing doc quality test?
#3529 kernel -> convolution, holding
#3532 write to ts. blocked for no reason. Next level on 24/10/22
#3575 panels->design, unblocked awaiting review

todo:

1. work out the exact process for conflict resolution in sktime. Is it documented anywhere
post PR a vote will happen, then start vote.
3. sort out #3528

### WP 1
1. NUMBA: see
https://hackmd.io/R3DY5YJQRQa1mT5XaZkXGQ
a. bebug n_jobs, try and scope and try David's fix.
b. float32/float64, chris suggested a solution. Meet tues or wed

2. Docs:

develop through https://github.com/MatthewMiddlehurst/sktime-mm
changes to rtd2 branch automatically rebuild
https://readthedocs.org/projects/sktime-mm/

3. Any new/outstanding issues to prioritise? None presently.

### WP 2
1. Regression. Port david's trello into https://trello.com/b/prjGEY8R/regression
Add Matthew.
2. Bake off. Track here
https://hackmd.io/FdV8mCZlRj2EGHoePBIvHA
and here
https://github.com/uea-machine-learning/estimator-evaluation/issues
check we can add in train files later

todo:
2.1 regression problems -> classification problems
2.2 collate HC2 in tsml with sktime files for multivariate
2.3 run current DL classifiers on GPU cluster
2.4 form new list for univariate and run missing HC2

Matthew to document whats running etc.
Tony to run Feature based ones.

3. Clustering. do comparison to tsml and sktime?

### WP3
EEG: Aiden starting properly this week.
Vitals: Meeting CHIMERA person at