Diego: Data IntElliGence Out.
Project description
Conan Reborn
Diego: Data IntElliGence Out. Sprite Come from Fast.ai and MicroSoft nni.
模块结构
study, trials
参考MicroSoft nni,定义Study
和Trial
。
每次的任务认为是一个Study
,每个 Study 由多个Trial
构成。
建议先创建 Study,再从 Study 中生成 Trial:
from diego.study import create_study
import sklearn.datasets
digits = sklearn.datasets.load_digits()
X_train, X_test, y_train, y_test = sklearn.model_selection.train_test_split(digits.data, digits.target,train_size=0.75, test_size=0.25)
s = create_study(X_train, y_train)
# can use default trials in Study
# or generate one
s.generate_trials(mode='fast')
s.optimize(X_test, y_test)
all_trials = s.get_all_trials()
for t in all_trials:
print(t.__dict__)
print(t.clf.score(X_test, y_test))
core
storage
对于每次的Study,数据的存储和参数,以及模型是额外存在Storage
对象的,保证了Study只控制trials,同时每个Trial完成后更新在storage中的结果,同时更新最好的结果。
结果的更新
在创建Study
的时候,需要指定优化的方向 maximize
或者 minimize
。同时在创建Trials
的时候,指定优化的指标。默认是 maximize accuracy
。
TODO 文档更新。
features TODO
- 不同类型的Trial。TPE, BayesOpt, RandomSearch
- 自定义的Trial。Trials by custom Classifier (like sklearn, xgboost)
- 模型保存。model persistence
- 模型输出。model output
- basic Classifier
- fix mac os hanged in optimize pipeline
train
用于组装数据、模型、损失函数、优化方法等。 定义:
class Learner():
"Trainer for `model` using `data` to minimize `loss_func` with optimizer `opt_func`."
class Recorder(LearnerCallback):
"A `LearnerCallback` that records epoch, loss, opt and metric data during training."
class Estimator():
"A estimator to evaulate learner. "
auto ml 补完计划
bayes opt
grid search
- H2O.ai
tree parzen
- hyperopt
- mlbox
metaheuristics grid search
- pybrain
generation
1.tpot
dl
- ms nni
installation
install swig
推荐使用 conda 安装
conda install swig
其他 dep
pip install pyrfr
pip install smac
pip install autosklearn
issues
updates
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
diego-0.1.1.tar.gz
(48.4 kB
view hashes)
Built Distribution
diego-0.1.1-py2.py3-none-any.whl
(59.5 kB
view hashes)
Close
Hashes for diego-0.1.1-py2.py3-none-any.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | 5db5b18f5a4b0a7e7e569dcc58e1465c4f18615ec470d5b51180ded5a7e2d016 |
|
MD5 | b041bd0c13de953269bd5c5fa017e5c1 |
|
BLAKE2b-256 | 6b97f5573cb51831c0025cd2cc2b5e98821cec5e8159b416ac668e7c02165543 |