profile
viewpoint
Sebastian Raschka rasbt UW-Madison Madison, WI http://sebastianraschka.com Machine Learning researcher & open source contributor. Author of "Python Machine Learning." Asst. Prof. of Statistics @ UW-Madison.

rasbt/deeplearning-models 11599

A collection of various deep learning architectures, models, and tips

rasbt/mlxtend 2769

A library of extension and helper modules for Python's data analysis and machine learning libraries.

rasbt/deep-learning-book 2641

Repository for "Introduction to Artificial Neural Networks and Deep Learning: A Practical Guide with Applications in Python"

rasbt/matplotlib-gallery 592

Examples of matplotlib codes and plots

amueller/scipy-2016-sklearn 492

Scikit-learn tutorial at SciPy2016

rasbt/algorithms_in_ipython_notebooks 418

A repository with IPython notebooks of algorithms implemented in Python.

rasbt/musicmood 369

A machine learning approach to classify songs by mood.

rasbt/biopandas 237

Working with molecular structures in pandas DataFrames

rasbt/datacollect 195

A collection of tools to collect and download various data.

Benjamin-Lee/deep-rules 168

Ten Quick Tips for Deep Learning in Biology

issue commentrasbt/python-machine-learning-book

IndexError: too many indices for array in CH-5 PCA Plot Code

There error means that X_train_pca has either 0 or 1 dimensions, instead of 2, or there is only 1 column instead of 2. You can double-check by executing print(X_train_pca.shape) before the for loop. The returned tuple should have at least 2 numbers, where the second number should be a 2.

This if you see that X_train_pca.shape is not (100, 2), then there is a typo somewhere in your preceding code.

somesh636

comment created time in a day

push eventrasbt/stat453-deep-learning-ss20

rasbt

commit sha 3b5af531d0ca14caabd115ef2ad7547641eae497

make L05 & L06 gradients in code more intuitive

view details

push time in a day

push eventrasbt/stat453-deep-learning-ss20

rasbt

commit sha 4399c216e1268b59034afb9dad808e0c854c634f

answers to some common L05 questions

view details

push time in 2 days

push eventrasbt/stat453-deep-learning-ss20

rasbt

commit sha 4631675118333afc43c6531a9d9b76eec64925a9

code for L06

view details

push time in 2 days

push eventconda-forge/mlxtend-feedstock

regro-cf-autotick-bot

commit sha 3ae26a0aa9d70bf60d88c7fddd2c23de941a26a9

use matplotlib-base instead of matplotlib

view details

regro-cf-autotick-bot

commit sha fe4060a76f62caee444df4bbb5d5cc20b4dc47ae

MNT: Re-rendered with conda-build 3.18.11, conda-smithy 3.6.8, and conda-forge-pinning 2020.02.12

view details

Sebastian Raschka

commit sha bcd3498b74fd7d2d906291c532e5171cbadc0e4b

Merge pull request #19 from regro-cf-autotick-bot/matplotlib-to-matplotlib-base-migration Suggestion: depend on matplotlib-base instead of matplotlib

view details

push time in 2 days

PR merged conda-forge/mlxtend-feedstock

Reviewers
Suggestion: depend on matplotlib-base instead of matplotlib

I noticed that this recipe depends on matplotlib instead of matplotlib-base. Unless you need pyqt, recipes should depend only on matplotlib-base. This PR makes this change. Notes and instructions for merging this PR:

  1. Make sure that the recipe can indeed only depend on matplotlib-base.
  2. Please merge the PR only after the tests have passed.
  3. Feel free to push to the bot's branch to update this PR if needed.

If this PR was opened in error or needs to be updated please add the bot-rerun label to this PR. The bot will close this PR and schedule another one. If you do not have permissions to add this label, you can use the phrase <code>@<space/>conda-forge-admin, please rerun bot</code> in a PR comment to have the conda-forge-admin add it for you.

<sub>This PR was created by the cf-regro-autotick-bot. The cf-regro-autotick-bot is a service to automatically track the dependency graph, migrate packages, and propose package version updates for conda-forge. If you would like a local version of this bot, you might consider using rever. Rever is a tool for automating software releases and forms the backbone of the bot's conda-forge PRing capability. Rever is both conda (conda install -c conda-forge rever) and pip (pip install re-ver) installable. Finally, feel free to drop us a line if there are any issues! This PR was generated by https://circleci.com/gh/regro/circle_worker/14377, please use this URL for debugging</sub>

+11 -2

1 comment

3 changed files

regro-cf-autotick-bot

pr closed time in 2 days

issue closedRaschka-research-group/coral-cnn

Questions on (rank-monotonic) and (num_classes - 1)

Hi, Thank you for your code implementation. I have three questions would like to ask:

  1. in the paper, it says "require {fk} reflect the ordinal information and are rank-monotonic". I am wondering how the rank monotonic is ensured?

  2. for the "levels" variable from your code, also the paper, why choose (num_classes - 1) instead of num_classes?

Thanks

closed time in 2 days

Shaunlipy

issue commentRaschka-research-group/coral-cnn

Questions on (rank-monotonic) and (num_classes - 1)

Hi Shaun,

in the paper, it says "require {fk} reflect the ordinal information and are rank-monotonic". I am wondering how the rank monotonic is ensured?

the rank monotonicity is guaranteed. The proof is provided in Theorem 1. (While not necessary, I also computed the violations of rank monotonicity, which in case of CORAL was 0, as expected. This is summarized in Table 2.)

for the "levels" variable from your code, also the paper, why choose (num_classes - 1) instead of num_classes?

Suppose you have 3 classes, 0, 1, 2. Then, if you know whether y > 2 , by knowing y > 0 and y > 1. Hence, the last one is redundant. E.g., if you know y>0 is True and y>1 is true, then you know y=2. Or if y > 1 is not true, you know y=2 is false

Hope that helps!

Shaunlipy

comment created time in 2 days

issue commentrasbt/mlxtend

About Saving SFS Intermediate State

Arg, that sounds frustrating. Sorry to hear that regarding the crashing. Do you know if this is due to some multiprocessing/joblib related issue. And do you have the error message by chance?

Right now, there is no way to save an intermediate state. The only thing I can imagine doing right now is running the SFS (via backward selection) for down to say 26 features, then safe/print the feature subset and use it in another round of SFS.

    import yaml

    sfs1 = SequentialFeatureSelector(..., k_features=26)
    sfs1.fit(X, y)
    print(sfs1.subsets_)
    yaml.dump(sfs1.subsets_, 'savefile.yaml')

Then in a new session

    import yaml

    previous_subsets = yaml.load('savefile.yaml')

    sfs2 = SequentialFeatureSelector(..., k_features=1)
    sfs2.fit(X[previous_subsets[26]], y)
    print(sfs2.subsets_)

Another way would be to add an optional checkpointing parameter that saves a pickle file of the current object in each iteration.

sskarkhanis

comment created time in 2 days

issue closedrasbt/python-machine-learning-book-3rd-edition

ch3,p56,“niter” parameter missed

with the initialized of the Perceptron class , the book mentioned the parameter "n_iter", the i cannot found in the code, and the code of 2nd has it. so i think it should be a issue QQ20200217-112213 QQ20200217-112200

closed time in 3 days

konfer

issue commentrasbt/python-machine-learning-book-3rd-edition

ch3,p56,“niter” parameter missed

Thanks for the note! You are right, I missed that part of the sentence during editing, and the n_iter parameter has been removed from sklearn in v 0.21. Just added it to the errata (https://github.com/rasbt/python-machine-learning-book-3rd-edition/blob/master/errata/README.md). Thanks!

konfer

comment created time in 3 days

push eventrasbt/python-machine-learning-book-3rd-edition

Sebastian Raschka

commit sha c55910f4188dc7977f2f08dba913b84c9476a23a

Update README.md

view details

push time in 3 days

issue closedrasbt/mlxtend

Does StackingRegressor split data for base model and meta model?

I didn't see the option to split the data when using StackingRegressor(), but as I learn for non-cross-validate stacking we split the data to two part, one for base model the other for meta-model training.

If the StackingRegressor() don't split the data, the meta-regressor receive predictions based on same data from base-model, than the whole ensemble may cause over-fitting problem.

closed time in 4 days

laurence-lin

issue commentrasbt/mlxtend

Does StackingRegressor split data for base model and meta model?

Yes, I agree that it can cause overfitting problems due to the reasons you mentioned. That's why I would recommend the StackingCVClassifier. If you provide your own split as cv argument, it would behave like you said, i.e, that the dataset is split into two but without cross-validation.

laurence-lin

comment created time in 4 days

issue commentrasbt/deeplearning-models

Imbalanced Classes

Was there maybe a formatting issue in the CSV file? Since you closed that issue, I assume you resolved it?

SURABHI-GUPTA

comment created time in 4 days

push eventrasbt/python-machine-learning-book-3rd-edition

Sebastian Raschka

commit sha 69b23772c2fc951e622302f9ac107a5130febd99

Update README.md

view details

push time in 4 days

push eventrasbt/python-machine-learning-book-3rd-edition

Sebastian Raschka

commit sha 9e6505d14d9980ff5ef74f7f417d139f839069b4

Update README.md

view details

push time in 4 days

push eventrasbt/stat453-deep-learning-ss20

rasbt

commit sha 6389082397785a5e6e52defa500f2ce36d6d246a

L06: Automatic differentiation & PyTorch

view details

push time in 7 days

push eventrasbt/stat453-deep-learning-ss20

rasbt

commit sha 08f77d38f48aa79a82eda9215897480ec1a50b9f

intalling pytorch

view details

push time in 7 days

push eventrasbt/stat453-deep-learning-ss20

rasbt

commit sha 14043daa7ebc682c1df10993b00a0a23d985a938

l05 code

view details

push time in 7 days

pull request commentrasbt/python-machine-learning-book-3rd-edition

Ch09 - Update the variable for the db connection in app.py

Thanks for the PR. Just wondering where in the code you faced issues? Afaik running the code with review_db didn't cause any problems in my case.

vidyabhandary

comment created time in 7 days

push eventrasbt/python-machine-learning-book-3rd-edition

natowi

commit sha 9dc8cf4a3713ddeeec86027b7a1f838bbfe9642a

Improve badge links

view details

Sebastian Raschka

commit sha 1af31f01696d40bd9dfe3df9551041fcfb7aa149

Merge pull request #113 from natowi/patch-1 Improve badge links

view details

push time in 7 days

pull request commentrasbt/python-machine-learning-book-3rd-edition

Improve badge links

Thanks!

natowi

comment created time in 7 days

push eventrasbt/stat453-deep-learning-ss20

rasbt

commit sha 29c2e59860b155f07360ba5e3edd456ff408a82f

L05: gradient descent

view details

push time in 9 days

push eventrasbt/stat453-deep-learning-ss20

rasbt

commit sha 7648e1e582fff91a00497b753c2e0d8e6c5f3034

examle proposal

view details

push time in 9 days

issue commentrasbt/pyprind

Not working in while statement

Hm, that's weird. Just tried your example and it worked. Did you maybe have a different indentation? Screen Shot 2020-02-10 at 2 58 00 PM

xyus

comment created time in 9 days

issue commentrasbt/mlxtend

stacking_cv_regression.py should use X, y = check_X_y(X, y, accept_sparse=['csc', 'csr'], dtype=None)

just catching up with notifications: Made a 0.17.1 version ~2 weeks back that should have that fix :).

tjhgit

comment created time in 9 days

push eventrasbt/python-machine-learning-book-3rd-edition

rasbt

commit sha 54088b27fcaed92f349e817c3f2dc03db0678bda

fix typo in nb 09 'we now save' -> 'we know save'

view details

Sebastian Raschka

commit sha d3baff98dc434e19c17b54617a8fc7bba36198f9

Merge pull request #111 from rasbt/typo09 fix typo in nb 09 'we know save' -> 'we now save'

view details

push time in 13 days

create barnchrasbt/python-machine-learning-book-3rd-edition

branch : typo09

created branch time in 13 days

push eventrasbt/stat453-deep-learning-ss20

rasbt

commit sha c4501fa99f8ba0f7aca9a4d1a073b4e276bf2d9e

add slides for L04

view details

push time in 16 days

issue closedrasbt/mlxtend

Get scores by feature

Hello,

I just started to use this package and I would like to know how to get the scores of the model by feature.

closed time in 16 days

shllgtca

issue commentrasbt/mlxtend

Get scores by feature

glad that i could clarify. please let me know in case you have any other questions about this

shllgtca

comment created time in 16 days

push eventrasbt/stat453-deep-learning-ss20

rasbt

commit sha 4dc6242533e94ca8e53c7a3dbfd096dda3605667

fix links

view details

push time in 20 days

push eventrasbt/stat453-deep-learning-ss20

rasbt

commit sha 8fa91e86c7d605d7f4f2bff76f34e1e2a40ca683

don't load class label into X array

view details

push time in 20 days

issue commentrasbt/mlxtend

Get scores by feature

I see. So in the SFS, the importance or score is not computed on a per-feature basis. What's computed is the "score" for a subset of features. The philosophy behind SFS is that it does not assume independence between features. I.e,. there could be features that are only important in combination with others.

That being said, let's say you selected a feature subset, you can analyze the individual performance of each feature in that subset using e.g., http://rasbt.github.io/mlxtend/user_guide/evaluate/feature_importance_permutation/ as a follow-up

shllgtca

comment created time in 20 days

create barnchrasbt/stat453-deep-learning-ss20

branch : numpy-vs-pytorch-compare-1

created branch time in 20 days

push eventrasbt/stat453-deep-learning-ss20

rasbt

commit sha 43009231eccd45424a017b126c2ac33d711f6269

making the pytorch code more numy like

view details

push time in 20 days

push eventrasbt/stat453-deep-learning-ss20

rasbt

commit sha 465e4377c93ea55a61907daa688b37ed8d2a1ca7

fix links

view details

push time in 20 days

push eventrasbt/stat453-deep-learning-ss20

rasbt

commit sha ee628fe313101de3d1ade059d77dc2c2e08fae13

HW1

view details

push time in 20 days

create barnchrasbt/stat453-deep-learning-ss20

branch : temp

created branch time in 21 days

push eventrasbt/stat453-deep-learning-ss20

rasbt

commit sha b70c84fda66e8d55c297ee0333c8f130001dd122

python installation help

view details

push time in 21 days

issue commentrasbt/mlxtend

Get scores by feature

Which model are you referring to, and what kind of scores are you after? In general, the following function will probably do what you want: http://rasbt.github.io/mlxtend/user_guide/evaluate/feature_importance_permutation/

shllgtca

comment created time in 21 days

push eventconda-forge/mlxtend-feedstock

regro-cf-autotick-bot

commit sha 95b9798c77a25ae7ea13d1f4b23b75d48598868b

updated v0.17.1

view details

regro-cf-autotick-bot

commit sha 0e36fb79c5c2d7acaab5e17adf255fae27bd30f8

MNT: Re-rendered with conda-build 3.18.11, conda-smithy 3.6.6, and conda-forge-pinning 2020.01.27

view details

Sebastian Raschka

commit sha 814df8389c6e0c6ed2109c1a12c2dd956691eb73

Merge pull request #18 from regro-cf-autotick-bot/0.17.1 mlxtend v0.17.1

view details

push time in 22 days

PR merged conda-forge/mlxtend-feedstock

Reviewers
mlxtend v0.17.1

It is very likely that the current package version for this feedstock is out of date. Notes for merging this PR:

  1. Feel free to push to the bot's branch to update this PR if needed.
  2. The bot will almost always only open one PR per version. Checklist before merging this PR:
  • [x] Dependencies have been updated if changed
  • [x] Tests have passed
  • [x] Updated license if changed and license_file is packaged

Note that the bot will stop issuing PRs if more than 3 Version bump PRs generated by the bot are open. If you don't want to package a particular version please close the PR.

If this PR was opened in error or needs to be updated please add the bot-rerun label to this PR. The bot will close this PR and schedule another one. If you do not have permissions to add this label, you can use the phrase <code>@<space/>conda-forge-admin, please rerun bot</code> in a PR comment to have the conda-forge-admin add it for you.

<sub>This PR was created by the cf-regro-autotick-bot. The cf-regro-autotick-bot is a service to automatically track the dependency graph, migrate packages, and propose package version updates for conda-forge. If you would like a local version of this bot, you might consider using rever. Rever is a tool for automating software releases and forms the backbone of the bot's conda-forge PRing capability. Rever is both conda (conda install -c conda-forge rever) and pip (pip install re-ver) installable. Finally, feel free to drop us a line if there are any issues! This PR was generated by https://circleci.com/gh/regro/circle_worker/13987, please use this URL for debugging</sub>

+36 -10

1 comment

7 changed files

regro-cf-autotick-bot

pr closed time in 22 days

push eventrasbt/mlxtend

Sebastian Raschka

commit sha 6c215b354051d92e3cbeaf4db3e39eb0bbe1a882

init setup for next version (#661)

view details

push time in 22 days

PR merged rasbt/mlxtend

init setup for next version
+26 -1

0 comment

2 changed files

rasbt

pr closed time in 22 days

pull request commentrasbt/watermark

handle tuple & list version numbers

Thanks! Do you know a specific library that uses this format? I think it would good to add this as an example to the docs for testing. Maybe it's about time to setup unittests as well -- if that's possible with jupyter notebooks.

lsternlicht

comment created time in 22 days

PR opened rasbt/mlxtend

init setup for next version
+26 -1

0 comment

2 changed files

pr created time in 22 days

create barnchrasbt/mlxtend

branch : v0.18.0dev

created branch time in 22 days

push eventrasbt/mlxtend

rasbt

commit sha ccfb58421166909231abe3fa9a3e70e875990f58

Deployed 8c7df5b6 with MkDocs version: 1.0.4

view details

push time in 22 days

created tagrasbt/mlxtend

tagv0.17.1

A library of extension and helper modules for Python's data analysis and machine learning libraries.

created time in 22 days

release rasbt/mlxtend

v0.17.1

released time in 22 days

push eventrasbt/mlxtend

Sebastian Raschka

commit sha 8c7df5b69ecfb39c948f7b64e1ab04aceaea0e35

fix download links

view details

push time in 22 days

push eventrasbt/mlxtend

Sebastian Raschka

commit sha 183f00e53209c843a275229791b14d08ae946b75

v0.17.1 (#660)

view details

push time in 22 days

PR merged rasbt/mlxtend

v0.17.1

Description

New version release

New Features
  • The SequentialFeatureSelector now supports using pre-specified feature sets via the fixed_features parameter. (#578)
  • Adds a new accuracy_score function to mlxtend.evaluate for computing basic classifcation accuracy, per-class accuracy, and average per-class accuracy. (#624 via Deepan Das)
  • StackingClassifier and StackingCVClassifiernow have a decision_function method, which serves as a preferred choice over predict_proba in calculating roc_auc and average_precision scores when the meta estimator is a linear model or support vector classifier. (#634 via Qiang Gu)
Changes
  • Improve the runtime performance for the apriori frequent itemset generating function when low_memory=True. Setting low_memory=False (default) is still faster for small itemsets, but low_memory=True can be much faster for large itemsets and requires less memory. Also, input validation for apriori, ̀ fpgrowthandfpmaxtakes a significant amount of time when input pandas DataFrame is large; this is now dramatically reduced when input contains boolean values (and not zeros/ones), which is the case when usingTransactionEncoder`. (#619 via Denis Barbier)
  • Add support for newer sparse pandas DataFrame for frequent itemset algorithms. Also, input validation for apriori, ̀ fpgrowthandfpmax` runs much faster on sparse DataFrame when input pandas DataFrame contains integer values. (#621 via Denis Barbier)
  • Let fpgrowth and fpmax directly work on sparse DataFrame, they were previously converted into dense Numpy arrays. (#622 via Denis Barbier)
Bug Fixes
  • Fixes a bug in mlxtend.plotting.plot_pca_correlation_graph that caused the explaind variances not summing up to 1. Also, improves the runtime performance of the correlation computation and adds a missing function argument for the explained variances (eigenvalues) if users provide their own principal components. (#593 via Gabriel Azevedo Ferreira)
  • Behavior of fpgrowth and apriori consistent for edgecases such as min_support=0. (#573 via Steve Harenberg)
  • fpmax returns an empty data frame now instead of raising an error if the frequent itemset set is empty. (#573 via Steve Harenberg)
  • Fixes and issue in mlxtend.plotting.plot_confusion_matrix, where the font-color choice for medium-dark cells was not ideal and hard to read. #588 via sohrabtowfighi)
  • The svd mode of mlxtend.feature_extraction.PrincipalComponentAnalysis now also n-1 degrees of freedom instead of n d.o.f. when computing the eigenvalues to match the behavior of eigen. #595
  • Disable input validation for StackingCVClassifier because it causes issues if pipelines are used as input. #606
+2 -2

0 comment

43 changed files

rasbt

pr closed time in 22 days

pull request commentrasbt/mlxtend

[WIP] Implement apriori-gen as in original paper

Sorry, still haven't had time to look into this more. Along with the new semester (lots of teaching) & 2 paper deadlines in January, there wasn't time for much else recently. I am currently making a 0.17.1 bugfix release with the recent changes -- someone from industry contacted me about this, because due to company firewalls, several people can only install it from PyPI (not from GitHub directly). Will revisit this PR soon though -- thanks for all the work on it so far!

dbarbier

comment created time in 22 days

IssuesEvent

issue commentrasbt/mlxtend

Plotting Apriori output with arulesViz

Thanks a lot @dbarbier ! I will leave this open to remind me to add this to the docs, because it's a nice addition that some people find useful for reference. I am planning to add this to the apriori docs and then linking it in the fgrowth and fpmax docs as well. Thanks!

archienorman11

comment created time in 22 days

PR opened rasbt/mlxtend

v0.17.1

Description

New version release

New Features
  • The SequentialFeatureSelector now supports using pre-specified feature sets via the fixed_features parameter. (#578)
  • Adds a new accuracy_score function to mlxtend.evaluate for computing basic classifcation accuracy, per-class accuracy, and average per-class accuracy. (#624 via Deepan Das)
  • StackingClassifier and StackingCVClassifiernow have a decision_function method, which serves as a preferred choice over predict_proba in calculating roc_auc and average_precision scores when the meta estimator is a linear model or support vector classifier. (#634 via Qiang Gu)
Changes
  • Improve the runtime performance for the apriori frequent itemset generating function when low_memory=True. Setting low_memory=False (default) is still faster for small itemsets, but low_memory=True can be much faster for large itemsets and requires less memory. Also, input validation for apriori, ̀ fpgrowthandfpmaxtakes a significant amount of time when input pandas DataFrame is large; this is now dramatically reduced when input contains boolean values (and not zeros/ones), which is the case when usingTransactionEncoder`. (#619 via Denis Barbier)
  • Add support for newer sparse pandas DataFrame for frequent itemset algorithms. Also, input validation for apriori, ̀ fpgrowthandfpmax` runs much faster on sparse DataFrame when input pandas DataFrame contains integer values. (#621 via Denis Barbier)
  • Let fpgrowth and fpmax directly work on sparse DataFrame, they were previously converted into dense Numpy arrays. (#622 via Denis Barbier)
Bug Fixes
  • Fixes a bug in mlxtend.plotting.plot_pca_correlation_graph that caused the explaind variances not summing up to 1. Also, improves the runtime performance of the correlation computation and adds a missing function argument for the explained variances (eigenvalues) if users provide their own principal components. (#593 via Gabriel Azevedo Ferreira)
  • Behavior of fpgrowth and apriori consistent for edgecases such as min_support=0. (#573 via Steve Harenberg)
  • fpmax returns an empty data frame now instead of raising an error if the frequent itemset set is empty. (#573 via Steve Harenberg)
  • Fixes and issue in mlxtend.plotting.plot_confusion_matrix, where the font-color choice for medium-dark cells was not ideal and hard to read. #588 via sohrabtowfighi)
  • The svd mode of mlxtend.feature_extraction.PrincipalComponentAnalysis now also n-1 degrees of freedom instead of n d.o.f. when computing the eigenvalues to match the behavior of eigen. #595
  • Disable input validation for StackingCVClassifier because it causes issues if pipelines are used as input. #606
+2 -2

0 comment

43 changed files

pr created time in 22 days

push eventrasbt/mlxtend

rasbt

commit sha 493bd9160f2c7567060e938323f021716baac983

v0.17.1

view details

push time in 22 days

create barnchrasbt/mlxtend

branch : v0.17.1

created branch time in 22 days

push eventrasbt/deeplearning-models

Sebastian Raschka

commit sha cd832fbd5a6f480e51b0df500ce60bb1f1f97d42

upd lenet models

view details

Sebastian Raschka

commit sha 6816908a3567ff7da539d9a2931047313c3f20dc

Merge pull request #39 from rasbt/lenet5 upd lenet models

view details

push time in 22 days

PR merged rasbt/deeplearning-models

upd lenet models

Fixes #38

+450 -438

1 comment

3 changed files

rasbt

pr closed time in 22 days

issue closedrasbt/deeplearning-models

a problem about LeNet-5 Classifier

Hi, The model of LeNet-5 contains no nonlinearity. Is it wrong?

closed time in 22 days

zhangruochi

PR opened rasbt/deeplearning-models

upd lenet models

Fixes #38

+450 -438

0 comment

3 changed files

pr created time in 22 days

create barnchrasbt/deeplearning-models

branch : lenet5

created branch time in 22 days

issue commentrasbt/deeplearning-models

a problem about LeNet-5 Classifier

oh wow, no that you mentioned that ... I remember having them in the forward method, not sure how that went missing. will fix that

zhangruochi

comment created time in 22 days

push eventrasbt/stat453-deep-learning-ss20

rasbt

commit sha 2394938ee42b3e593d7bc07f69b9c60b14ea74a8

L03: Perceptron

view details

push time in 23 days

push eventrasbt/stat453-deep-learning-ss20

rasbt

commit sha bfa08b2b859b9fa90b163987e9bf0cf415ad761c

perceptron animation

view details

rasbt

commit sha 11e07f7451c9d0761b3a7e31e323cca8680e1a4c

Merge branch 'master' of https://github.com/rasbt/stat453-deep-learning-ss20

view details

push time in 23 days

issue commentrasbt/deeplearning-models

Imbalanced Classes

The imbalances we were dealing with weren't that extreme so in our case oversampling and/or undersampling both worked fine. Not sure if that's true in your case though.

You can maybe pre-train the network on a different attribute that is less imbalanced. Then, you can undersample your current dataset and use transfer learning to see if you get good results. It's a big imbalance though and getting good results may be tricky.

SURABHI-GUPTA

comment created time in 23 days

issue commentrasbt/deeplearning-models

Imbalanced Classes

In that case, it could indeed have learned to always predict the majority label, because

19049/(19049+913) = 0.954

SURABHI-GUPTA

comment created time in 23 days

issue commentrasbt/deeplearning-models

Imbalanced Classes

it depends. On which labels (imbalanced features) did you do the prediction? And what's their ratio in the dataset?

SURABHI-GUPTA

comment created time in 23 days

issue commentrasbt/deeplearning-models

Imbalanced Classes

Sry we haven't done that, we only oversampled by gender and skin color

SURABHI-GUPTA

comment created time in 23 days

issue closedrasbt/deeplearning-models

Imbalanced Classes

Hi, Does this implementation solves any class imbalanced datasets ? Since of the features of Celeba dataset has high class imbalance, does your solution addresses this issue ?

closed time in 23 days

SURABHI-GUPTA

issue commentrasbt/deeplearning-models

Imbalanced Classes

Hi there,

good question! We did account for class imbalance in a research project where we used CelebA (https://arxiv.org/abs/1807.11936), but I didn't do anything regarding class imbalance in this repository.

This was because in an earlier project where we developed a method for "fooling" a gender classifier, I found that the gender classifier can more easily be fooled if it is a picture of a male person. It may be because there are more female pictures in the dataset. Regarding skin color and gender, there was also some difference for males; however, I didn't see the difference they found in the "Gender Shades" (http://gendershades.org/overview.html) paper, which was published a few months afterwards. Anyways, here are some of the results I got back then ~2 years ago. This is for fooling the gender classifier, so an error of 0.5 is desired.

Screen Shot 2020-01-27 at 10 16 53 AM

SURABHI-GUPTA

comment created time in 23 days

issue commentrasbt/mlxtend

StackingCVClassifier: Passing parameters to the fit method of the underlying classifiers

I am currently a bit swamped -- the new semester just started plus I have a paper due next week, but I was planning to make a new mlxtend release this week (or very soon). The reason is that there were some bugs in the previous release, and there are some people in industry who can only install PyPI packages (not from GitHub directly) due to company firewalls.

To keep a long story short, if you are willing to contribute with a PR, that would be very appreciated!

MiladShahidi

comment created time in 25 days

issue closedrasbt/mlxtend

Comparison with sklearn's StackingClassifier

Apologies if this is not the place for such a question, but:

What are the differences between mlxtend's StackingCVClassifier and sklearn's new StackingClassifier (which has a cv option)?

closed time in a month

rchaves33

issue commentrasbt/mlxtend

Comparison with sklearn's StackingClassifier

Good question. The StackingClassifier and StackingCVClassifier in mlxtend have been around for a long time (a few years) and I assume that the new scikit-learn StackingClassifier implements the same general approach but with a slightly different API. I haven't looked at scikit-learn's stacking classifier in detail though and can't tell you much more about it.

However, when I look at the documentation, it seem that it is using cross-validation internally (the recommended approach), so it is more similar to the StackingCVClassifier in mlxtend rather then the regular StackingClassifier in mlxtend.

I hope that helps!

rchaves33

comment created time in a month

push eventrasbt/stat453-deep-learning-ss20

rasbt

commit sha c9fb684d6c1bc7775664ffb43daf2fd2f13253f6

fix link to reading material in L01

view details

push time in a month

push eventrasbt/stat453-deep-learning-ss20

rasbt

commit sha 2370cd87132f86373c9e64ae419588fa0c9aae3c

L02

view details

rasbt

commit sha 0e74101fc7ad026fdd52589d81bc4587bcea6193

Merge branch 'master' of https://github.com/rasbt/stat453-deep-learning-ss20

view details

rasbt

commit sha 46cbda77675ca47095166b97f2fd11f03447bb93

upd gitignore

view details

push time in a month

issue commentRaschka-research-group/coral-cnn

While traning my model i'm facing issue.

I don't think this issue is related to CORAL; you may have some issues with your dataset because it cannot complete the first epoch. I suggest you iterate over your custom dataset and check the tensor sizes of features and targets to see if they are inconsistent somewhere.

NarasimmanSaravana1994

comment created time in a month

issue commentrasbt/mlxtend

Apriori - I'm only interested in one product / consequent. Can I speed up the apriori algorithm?

Sorry, there's currently no option for that. A "must_contain=('some item', 'some other item', ...)" feature could be a useful enhancement though.

soliverc

comment created time in a month

push eventrasbt/python-machine-learning-book-3rd-edition

Sebastian Raschka

commit sha 507254fc2d3bf0309426cbbcda59eb3d3695de13

Update README.md

view details

push time in a month

issue commentrasbt/mlxtend

Plotting Apriori output with arulesViz

Haven't tried it personally. The 2D versions look like sth that could be readily done in matplotlib. The interactive aspects sound interesting though. Regarding plotting experience and tips for association rules, maybe @dbarbier has some pointers.

archienorman11

comment created time in a month

push eventrasbt/stat453-deep-learning-ss20

rasbt

commit sha fae024af509ef111119dfacc7890ca5ee8051a85

updated office hours

view details

push time in a month

push eventrasbt/stat453-deep-learning-ss20

rasbt

commit sha 655f2e92dfecb4e64af912dc6264667df4ac21b2

add L01

view details

push time in a month

push eventrasbt/stat453-deep-learning-ss20

rasbt

commit sha facb39dd422b98d5c6eb97a045fdf47b6b293dd7

add report template

view details

push time in a month

push eventrasbt/stat453-deep-learning-ss20

Sebastian Raschka

commit sha ecf94e50446dfff8df035ca7a5104979b9734afc

Update README.md

view details

push time in a month

create barnchrasbt/stat453-deep-learning-ss20

branch : master

created branch time in a month

created repositoryrasbt/stat453-deep-learning-ss20

Intro to Deep Learning and Generative Models @ UW-Madison

created time in a month

issue commentrasbt/biopandas

Renumbering residues

Sounds good, I agree. I am currently caught up with a pretty long to do list of other things (and the semester is going to start Tue); so I am not sure when I will get to this, yet. If someone wants to take a crack at it, I welcome PRs.

ajasja

comment created time in a month

issue commentRaschka-research-group/coral-cnn

Why the model cannot predict less than 14 year of age using Single Image Predictions?

Yeah, I trimmed the dataset because some of the classes had a very low number of examples per age class. E.g., only 1-3 pictures per age class (but I don't remember the exact number, just that I trimmed it because the number of examples was very low).

QaisarRajput

comment created time in a month

issue closedRaschka-research-group/coral-cnn

Why the model cannot predict less than 14 year of age using Single Image Predictions?

Its not an issue, more of a query. I see that in the Labels and train/test splits section it is mentioned that

UTKFace: labels 0-39 correspond to ages 21-60

But as we know UTKFace data has samples from age (0 to 116) UTKFACE.

Are these models not trained on the full data-sets because of training resource needed or is there another reason?

closed time in a month

QaisarRajput

push eventBenjamin-Lee/deep-rules

Benjamin Lee

commit sha bbb50001fc7ec8f2aa8064ac96e5fcff59b96e15

Reorder the sentences in the privacy tip

view details

Benjamin Lee

commit sha e1d3886b4e89eddeb2e9abc30f32b3a0e8dc7dd7

Remove an extraneous word

view details

Benjamin Lee

commit sha 7052204f52a93ada39ba7f845a09b5d2218308a4

Remove extra hyphens

view details

Sebastian Raschka

commit sha 9522ea5b6db67fd06f3a98ffd4c7ce518e7967aa

Merge pull request #207 from Benjamin-Lee/textlinting Just a few nitpicks upon a fresh reread of the paper

view details

push time in a month

PR merged Benjamin-Lee/deep-rules

Just a few nitpicks upon a fresh reread of the paper

Did you add yourself as a contributor if this is your first contribution?

<!-- Put an x in between the brackets to show you did! -->

  • [x] Yes, I added myself or am already a contributor

Any more details?

Hyphens, extra words, etc

+13 -11

0 comment

5 changed files

Benjamin-Lee

pr closed time in a month

more