Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

PR v3.6.0 #314

Merged
merged 62 commits into from
Feb 22, 2022
Merged

PR v3.6.0 #314

merged 62 commits into from
Feb 22, 2022

Conversation

gufengzhou
Copy link
Contributor

@gufengzhou gufengzhou commented Feb 22, 2022 •
edited
Loading

  • New default hyperparameter lambda that finds MOO-optimal lambda and thus removes the need of manually lamdba definition
  • New optional hyperparameter penalty.factor that further extends hyperparameter spaces and thus potentially better fit
  • New optimisation convergence rules & plots for each objective function telling if a model run has converged or not (NRMSE, DECOMP.RSSD. MAPE if calibrated)
  • Improved response function now also returns the response for exposure metrics (response on imps, GRP, newsletter sendings etc) and plots. Note that argument names and output class has changed. See demo.R for more details.
  • More budget allocation stability by defaulting fitting media variables from paid_media_vars to paid_media_spends. Spend exposure fitting with Michaelis Menten function will only serve robyn_response function output and plotting. robyn_allocator() now only relies on direct spend - response transformation.
  • Default beta coefficient signs: positive for paid & organic media and unconstrained for the rest. Users can still set signs manually.

laresbernardo and others added 30 commits February 7, 2022 10:33
@laresbernardo
feat: add lambda into nevergrad optimization
956344d
@laresbernardo
fix: no need to use new objective slot
c85a2d4
@laresbernardo
Force R >= 4.0
a7eb623
@laresbernardo
feat: new use_penalty_factor param
be74de3
@laresbernardo
fix: when lambda and penalty_* are passed, keep them
e319f16
@laresbernardo
fix: simplified fixed hyper-parameters logic
852fc31
@laresbernardo
feat: enabled fixed hyper-parameters by setting one value
317374b
@gufengzhou
corrected hyper_collect logic
53afa73
@laresbernardo
fix: mistakenly posted demo results
1f90951
@laresbernardo
fix: fixed and nono-fixed response curve working
5cd2c7b
@laresbernardo
docs: update documentations
a98ca60
@laresbernardo
feat: new check_allocator()
bd12367
@laresbernardo
fix: no warnings/errors on check() + nloptr and allocator breakdown
0a3711c
@laresbernardo
refactor: split even more robyn_allocator()
b47d0ac
@gufengzhou
fix: lm condition bug and missing alphas
5dbb3e8
@gufengzhou
refactor: adapt variable and parameter names for more intuition
5afd30f
@laresbernardo
docs: add_penalty_factor renaming
1c6ca4f
@laresbernardo
fix: enabled 42 colours on robyn_palette() #306
ed1e7eb
@gufengzhou
fix: adapt fixed hp for refresh
dafcdb1
@gufengzhou
fix: multiple cluster input when refreshing
af215b4
@laresbernardo
feat: new check_conv_error() on robyn_run() based on error's sd on la…
4c618bf 
…st quantile (default = 10)
@laresbernardo
fix: changed normalized ROIs into real ROIs on clustering plot
49c3b27
@gufengzhou
fix: adapt refresh with new hyperparameters & unify lambda_max in model
92a8413
@laresbernardo
fix: robyn_pareto()'s rbindlist for OutputModels
178a124
@laresbernardo
fix: change seed each trial by summing 1
d703cba
@laresbernardo
feat: frienly informative print method for robyn_models objects
0392cd5
@laresbernardo
feat: new friendly and informative print method for robyn_inputs()
a16405b
@laresbernardo
feat: new print.robyn_outputs method
9d7842c
@laresbernardo
feat: new print.robyn_allocator method
05fe27e
@gufengzhou
fix: refresh bug of hyper_updated location
e0f0331
gufengzhou and others added 26 commits February 16, 2022 23:35
@gufengzhou
refactor: spend default - adapted input & run functions
fbd8f3d
@gufengzhou
refactor: improve convergence plot with 3 dimensions
0039d58
@gufengzhou
refactor: set paid_media_spends as default instead of paid_media_vars
345ceae
@laresbernardo
fix: no globalVars on ggplot2
b0e21e7
@gufengzhou
doc: update demo.R and documentation
528f7a9
@gufengzhou
refactor: re-add lambda_control to avoid pipeline break & minor fix
4d461da
@gufengzhou
fix: inherit add_penalty_factor
b531611
@gufengzhou
doc: add new robyn_response docu on demo.R and function doc
4bfcdae
@laresbernardo
Merge branch 'lares' of https://github.com/facebookexperimental/Robyn …
8920db4 
…into lares
@laresbernardo
merge: G & B's commits
54dafce 
Merge branch 'lares' of https://github.com/facebookexperimental/Robyn into lares

# Conflicts:
#	demo/demo.R
@laresbernardo
fix: passing parameters from input to run
42555a3
@laresbernardo
feat: add scenario in allocation plots to avoid overwritting
686bb3e
@laresbernardo
feat: new optimal_response() function based on robyn_allocation() res…
c74dcec 
…ults
@laresbernardo
feat: new print.robyn_refresh method
5a5982b
@laresbernardo
fix: logic on not provided values in robyn_response
b5796dd
@gufengzhou
refactor: rename robyn_converge & export it, recover QA section in demo
fbddfe5 
- reason for recovering QA response is that we want to be less black-boxy and letusers to know how to use output lists and how to play with response function.
@gufengzhou
fix: various workflow bugs
eb9e13f 
- set convergence default to 20 cuts, having 5% as convergence quantile
- check iteration only for initial build
- print csv also for fixed workflow
- various refresh bugs
@gufengzhou
fix: run check
83da495
@laresbernardo
fix: hyper-parameters plot and robyn_response()'s return as list
9300149
@laresbernardo
docs: fix nloptr vignette URL #313
a0900f4
@gufengzhou
doc: guide & docu adapted
29a2fee
@laresbernardo
fix: enable legacy inputs and warn user
b08b31e
@gufengzhou
fix: hypersampling plot lambda range
5197df3
@laresbernardo
feat: double convergence criteria
1f03030
@laresbernardo
feat: enable robyn_convergence() params on robyn_run() + better docs …
50f559b 
…re criteria
@gufengzhou
refactor: adapted to dynamic convergence rules
95091c9
@facebook-github-bot facebook-github-bot added the CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. label Feb 22, 2022
@gufengzhou gufengzhou merged commit 68ecfc9 into main Feb 22, 2022
@laresbernardo laresbernardo deleted the lares branch February 22, 2022 16:43
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers
No reviews
Assignees
No one assigned
Labels
CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
4 participants
@gufengzhou @laresbernardo @sahbakn @facebook-github-bot