# Nov 1st aeon Module Meeting (Transformations) - Nov 1st, 2024
###### tags: `aeon-module-meeting`

#### 1st Nov 25, 2024 @ 12:00PM UTC
#### Previous meeting: N/A

Present:

### Bugs, Issues and PRs

#### 3 Bugs:
https://github.com/aeon-toolkit/aeon/issues/324

close and reopen

https://github.com/aeon-toolkit/aeon/issues/2210

More a testing/data types issue, revisit when we faciliate dense data structure return

https://github.com/aeon-toolkit/aeon/issues/2268
(also https://github.com/aeon-toolkit/aeon/issues/1840)
fixed by
https://github.com/aeon-toolkit/aeon/pull/2274
Hadi to review

#### 14 other open issues
Shapelets:
actually have these all implemented in tsml-eval: Tony to do
https://github.com/aeon-toolkit/aeon/issues/186

MP: four variants (see comment)
https://github.com/aeon-toolkit/aeon/issues/833

To do:
1. benchmark for time different implementations
2. merge transformers into one, possibly with am
3. Option for bespoke implementation
4. similarity search to either use bespoke or stumpy

RDST: use online. Not sure there is any benefit from using online CLOSED
https://github.com/aeon-toolkit/aeon/issues/894

Coverage for SFA: still at 72%
https://github.com/aeon-toolkit/aeon/issues/988
The whole SFA/SFAFast division needs resolving. Suggested plan

Combine SFA/SFAFast collection transformer
Adjust TDE to use new SFA

Binary IG issue: leave for now
https://github.com/aeon-toolkit/aeon/issues/1322

Rocket GPU to regressors. Assigned to Jonathan
https://github.com/aeon-toolkit/aeon/issues/1542

Differencer series transformer. Leave for now
https://github.com/aeon-toolkit/aeon/issues/1553

Tidy up rockets: should revisit before Oz. Maybe get round to it some time.
https://github.com/aeon-toolkit/aeon/issues/1699

pycatch22/catch22:
https://github.com/aeon-toolkit/aeon/issues/1937

Matthew to take a look

duplicate normalise/standard scaler
https://github.com/aeon-toolkit/aeon/issues/2075

Needs looking at

Smoothing notebook: cyril self assigned
https://github.com/aeon-toolkit/aeon/issues/2267

implementing algos
https://github.com/aeon-toolkit/aeon/issues/797
https://github.com/aeon-toolkit/aeon/issues/1480
https://github.com/aeon-toolkit/aeon/issues/1481

### Open PRs: 8
in reverse chronological order
1. New notebook (tony)
https://github.com/aeon-toolkit/aeon/pull/2274
Hadi to review
2. similarity search (antoine) in progress
https://github.com/aeon-toolkit/aeon/pull/2176

3. SAST Ivan: merged
https://github.com/aeon-toolkit/aeon/pull/2006

4. Catch22 Danieli: merged
https://github.com/aeon-toolkit/aeon/pull/1979

5. SFA Testing (draft) Matthew
https://github.com/aeon-toolkit/aeon/pull/1884

6. Minirocket with unequal length (Tony)
https://github.com/aeon-toolkit/aeon/pull/1746

7. Test pycatch22 (Hadi): Closed
https://github.com/aeon-toolkit/aeon/pull/1235

Gave rise to the new issue:

Issue relates to normalised series create an issue describing and possible solutions

8. Inverse transform for standard scalar (David)
https://github.com/aeon-toolkit/aeon/pull/1210
wrap this into the whole normaliser/scalar

TO DO

- [x] **Scaler/normalise etc**

https://github.com/aeon-toolkit/aeon/pull/1210
https://github.com/aeon-toolkit/aeon/issues/2075

start communication on the solution, end up with one class, make invertible, close 1210
See https://github.com/aeon-toolkit/aeon/pull/2289

- [ ] **Catch22**

Sort out what is going on with normalisation, create issue, decide how to resolve

- [ ] **ROCKET**

Tony to have another go

- [ ] **Shapelets**

Tony to have another go

- [ ] **Sort out transformer output types**
Decide on types
Collection of series
Single series
(Sparse) Table
Dense table

- [ ] Matrix profile
Decide on which to use

## Future direction

1. Notebooks
2. tags (tag as series-to-series)?
3. Output types: list/test/restrict? unified dictionary output?
4. Duplicate collection and series transformers. Do we need them?
5. statsmodels: could try remove dependency?

No strong feelings, Tony might have a go at this


AOB:
None