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

[WIP] Inputsize invariant dev #308

Open
wants to merge 16 commits into
base: master
Choose a base branch
from
Open

[WIP] Inputsize invariant dev #308

wants to merge 16 commits into from

Conversation

corochann
Copy link
Member

fixing relgcn, relgat calculation as well, due to ggnnpreprocessor change.

@corochann
Copy link
Member Author

performance comparison with tox21

PR #241 : v0.4.0
eval_results_tox21

PR #302 : after module refactoring, before input size invariant.

eval_results_tox21

This PR #308 : input size invariant
eval_results_tox21

@corochann
Copy link
Member Author

corochann commented Jan 23, 2019 •
edited

performance comparison with qm9

It is difficult to see the difference after scaling...

In [1]: scaler.mean_
Out[1]:
array([ 9.81423703e+00,  1.40609910e+00,  1.12491609e+00,  2.70606047e+00,
        7.51916720e+01, -2.39976429e-01,  1.11237232e-02,  2.51099679e-01,
        1.18952558e+03,  1.48523891e-01, -4.11610053e+02, -4.11594996e+02,
       -4.11593263e+02, -4.11665414e+02,  3.16009336e+01])

In [2]: scaler.var_
Out[2]:
array([3.27409900e+06, 2.50771379e+00, 1.20007181e+00, 2.34202719e+00,
       6.70371780e+01, 4.89749655e-04, 2.20295275e-03, 2.25798483e-03,
       7.82604766e+04, 1.10710843e-03, 1.60476355e+03, 1.60471277e+03,
       1.60470911e+03, 1.60495593e+03, 1.65031147e+01])

In [3]: scaler.scale_
Out[3]:
array([1.80944715e+03, 1.58357627e+00, 1.09547789e+00, 1.53036832e+00,
       8.18762347e+00, 2.21302882e-02, 4.69356234e-02, 4.75182579e-02,
       2.79750740e+02, 3.32732389e-02, 4.00595001e+01, 4.00588663e+01,
       4.00588206e+01, 4.00619013e+01, 4.06240258e+00])

PR #241 : v0.4.0
eval_mean_abs_error_qm9

PR #302 : after module refactoring, before input size invariant.
eval_mean_abs_error_qm9

This PR #308 : input size invariant
eval_mean_abs_error_qm9

It seems metric for how to scale mean_abs_error has changed... I need to investigate.

@codecov-io
Copy link

codecov-io commented Jan 23, 2019 •
edited

Codecov Report

Merging #308 into master will increase coverage by 0.29%.
The diff coverage is 97.95%.

@@            Coverage Diff             @@
##           master     #308      +/-   ##
==========================================
+ Coverage   83.66%   83.95%   +0.29%     
==========================================
  Files         179      179              
  Lines        8098     8171      +73     
==========================================
+ Hits         6775     6860      +85     
+ Misses       1323     1311      -12

@corochann corochann mentioned this pull request Jan 23, 2019
Closed
corochann added 2 commits January 23, 2019 14:08
Merge branch 'master' into inputsize_invariant_ggnn2
cb577d2 
# Conflicts:
#	chainer_chemistry/links/readout/ggnn_readout.py
Merge branch 'inputsize_invariant_ggnn2' into inputsize_invariant_dev
e5fffb5 
# Conflicts:
#	chainer_chemistry/links/readout/ggnn_readout.py
@mottodora mottodora added this to the 0.6.0 milestone Feb 7, 2019
@mottodora mottodora modified the milestones: 0.6.0, 0.7.0 Sep 11, 2019
@corochann corochann removed this from the 0.7.0 milestone Dec 9, 2019
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
no-compat
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.

None yet
3 participants
@corochann @codecov-io @mottodora