Generalize (F)(L)ODOG normalization #3

JorisVincent · 2024-03-04T17:27:16Z

Documentation outlining generalization of (F)(L)ODOG normalizations
Refactor normalization-module, and subsequently of models-module
Test F,L,ODOG against MATLAB: passing
- Automate testing using GHA, Nox

Also properly document `padval` default

Where possible compare components to MATLAB output; else, compare to RHS transplementation

Easier to test against subcomponents of LODOG in transplementation

(Rather than no padding)

... for example stimulus, replicates exactly what the MATLAB code produces Merge branch 'dev_LODOG' into dev

extract normalizers to subroutine reuse LODOG gaussian kernel separate spatial masks reuse LODOG RMS function extract local RMS subroutine separate read out from normalization default spatial averaging mask is circular rename fixture

…xtures

as `.py`-script

Brings more in line with other doc pages

(and pseudo-implementation)

Reorganizes and removes some existing normalization Better specification of shape for spatial normalization

Closes #3 (Gitlab)

Since it is used to prevent DivideByZero errors, makes sense to use it in divisive_normalization() Default eps=0.0 Takes on different values in F/L/ODOG tests

Closes #4

feat(models): `norm_energies` method to replace `normalizers_to_RMS`

Closes #5

JorisVincent and others added 30 commits October 24, 2023 17:42

test(transplementation): rename

fc3d67c

test(transplementation): explicit testing of transplementation

86c1866

test(transplementation): lightly refactor

0989210

Merge branch 'test_transplementation' into dev

0a42056

test: remove plotting from tests

8e1e193

test((F)LODOG): Fixtures for MATLAB output

c130c51

test: separate out fixtures for comparing to MATLAB output

45b6f5b

test(transplementation): filterbank, apply, use extracted fixtures

b9d7c96

test(transplementation): models use extracted fixtures

ef09683

Merge branch 'test_fixtures' into dev

e860a36

feat(filters,filterbank): .apply()-methods take padval-arg

f98655e

Also properly document `padval` default

test(filters): test padval

456ea76

test(transplementation): padval

67ef44e

Merge branch 'feat_padval' into dev

8b2cb38

test(transplementation): weight

182b3e9

test(ODOG): regression ODOG components

51412aa

Where possible compare components to MATLAB output; else, compare to RHS transplementation

test(ODOG): fix use renamed fixtures

25b807c

Merge branch 'test_ODOG' into dev

eca3b24

test(LODOG, transplement): LODOG normalization

e389437

Easier to test against subcomponents of LODOG in transplementation

test(transplementation): test LODOG transplementation

b8f2b2a

test(LODOG): against transplement components

88a88c6

feat(LODOG): window_sigma as constructor param

cba9122

fix(LODOG) additional epsilon offset

80a51bd

fix(LODOG): pad normalizer filtering with 0s

bb2e1a7

(Rather than no padding)

test(LODOG): slightly rename fixtures

e561811

Merge branch 'test_LODOG_components' into dev_LODOG

698575f

fix(LODOG): exactly replicates original Robinson, Hammon, de Sa (2007)

f28a8f1

... for example stimulus, replicates exactly what the MATLAB code produces Merge branch 'dev_LODOG' into dev

feat(FLODOG): model params in object initialization

3e5a3f5

test(transplement): FLODOG normalization

56958d8

extract normalizers to subroutine reuse LODOG gaussian kernel separate spatial masks reuse LODOG RMS function extract local RMS subroutine separate read out from normalization default spatial averaging mask is circular rename fixture

test((F)LODOG): normalization params (F)LODOG for MATLAB output as fi…

d60a4a4

…xtures

MiaSelene and others added 25 commits February 15, 2024 17:07

docs(normalization): explanatory text towards filter averaging

c46dd6b

docs(normalization): move & rename as .py file

1ab5d4e

docs(normalization): remove execution info

3dd09fd

docs(normalization): cleanup generalization of normalization step

5e1c6ca

as `.py`-script

docs(normalization): introduction to generalization

1a94049

Brings more in line with other doc pages

docs(normalization): start with formalization of normalization step

d0e3659

docs(normalization): comparison of interaction weights

7e9d707

docs(normalization): ODOG normalization divides by matrices

113ee12

docs(normalization): global image averaging as filter operation

9f41637

docs(normalization): summarize generalized normalization

f696fe5

(and pseudo-implementation)

docs(normalization): typesetting

ff06d7e

docs: add generalized normalization to TOC

084ab92

Merge branch 'docs/normalization' into dev/general

b4a793e

feat(normalization): generalized normalization functions

e5effee

Reorganizes and removes some existing normalization Better specification of shape for spatial normalization

test: normalization

e3029e7

Closes #3 (Gitlab)

fix(normalization): offset epsilon to avoid 0s in division, sqrt

fc67d21

Since it is used to prevent DivideByZero errors, makes sense to use it in divisive_normalization() Default eps=0.0 Takes on different values in F/L/ODOG tests

refactor(normalization): more consistently name normalization components

b3751d4

Closes #4

refactor(normalization): normalizing coefficients in single tensor dot

974bcdd

refactor(models): use generalized normalization

8f53908

feat(models): `norm_energies` method to replace `normalizers_to_RMS`

test(models): update to use generalized normalization

dff5bad

docs(normalization): use eps

aeffbcf

docs(normalization): clarify norm_coeffs LODOG == ODOG

2ad0cbe

docs(normalization): ODOG energies

26e9192

docs(models): docstrings, type annotations

64865ce

Closes #5

Merge branch 'dev/normalization' into dev/general

75235e0

JorisVincent added bug Something isn't working documentation Improvements or additions to documentation enhancement New feature or request labels Mar 4, 2024

JorisVincent self-assigned this Mar 4, 2024

JorisVincent force-pushed the dev/general branch from 56265e2 to 75235e0 Compare March 4, 2024 19:22

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Generalize (F)(L)ODOG normalization #3

Generalize (F)(L)ODOG normalization #3

JorisVincent commented Mar 4, 2024 •

edited

Loading

Generalize (F)(L)ODOG normalization #3

Are you sure you want to change the base?

Generalize (F)(L)ODOG normalization #3

Conversation

JorisVincent commented Mar 4, 2024 • edited Loading

JorisVincent commented Mar 4, 2024 •

edited

Loading