1st place submission to the AutoML competition - phase 2
Python TeX
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Failed to load latest commit information.
automl_lib Code dump Sep 15, 2015
papers/workshop Small changes Sep 24, 2015
.gitignore Startinga workshop paper Sep 17, 2015
LICENSE Update LICENSE Sep 15, 2015
README.md Update README.md Sep 15, 2015
agent.py Code dump Sep 15, 2015
automl.py Code dump Sep 15, 2015
constants.py Code dump Sep 15, 2015
data_management.py Code dump Sep 15, 2015
deep_sandpit.py Code dump Sep 15, 2015
demo.py Code dump Sep 15, 2015
dummy_decision_maker.py Code dump Sep 15, 2015
experiment.py Code dump Sep 15, 2015
experiments.py Code dump Sep 15, 2015
freezethaw.py Code dump Sep 15, 2015
global_data.py Code dump Sep 15, 2015
learners.py Code dump Sep 15, 2015
libscores.py Code dump Sep 15, 2015
managers.py Code dump Sep 15, 2015
metadata Code dump Sep 15, 2015
metalearners.py Code dump Sep 15, 2015
postprocessing.py Code dump Sep 15, 2015
run.py Code dump Sep 15, 2015
sandpit.py Code dump Sep 15, 2015
sandpit_two.py Code dump Sep 15, 2015
score.py Code dump Sep 15, 2015
stackcombiner.py Code dump Sep 15, 2015
util.py Code dump Sep 15, 2015

README.md

automl-phase-2

1st place submission to the AutoML competition - phase 2

A reduced implementation of freeze-thaw Bayesian optimization extended to choose computations in the context of ensemble construction via stacking. Base estimators include most things in SKLearn.

Code is mixed quality. We made an architectural choice of message passing early on - it has some nice properties but makes it difficult to understand and debug.