Skip to main content

An General Automated Machine Learning Framework

Project description

Hypernets

Python Versions TensorFlow Versions Downloads PyPI Version

Hypernets: A General Automated Machine Learning Framework

Hypernets is a general AutoML framework, based on which it can implement automatic optimization tools for various machine learning frameworks and libraries, including deep learning frameworks such as tensorflow, keras, pytorch, and machine learning libraries like sklearn, lightgbm, xgboost, etc. We introduced an abstract search space representation, taking into account the requirements of hyperparameter optimization and neural architecture search(NAS), making Hypernets a general framework that can adapt to various automated machine learning needs.

Overview

Conceptual Model

Illustration of the Search Space

Installation

pip install hypernets

Verify installation:

python -c "from examples import smoke_testing;"

Hypernets related projects

  • HyperGBM: A full pipeline AutoML tool integrated various GBM models.
  • HyperDT/DeepTables: An AutoDL tool for tabular data.
  • HyperKeras: An AutoDL tool for Neural Architecture Search and Hyperparameter Optimization on Tensorflow and Keras.
  • Cooka: Lightweight interactive AutoML system.
  • Hypernets: A general automated machine learning framework.

DataCanvas AutoML Toolkit

Neural Architecture Search

Examples

HyperKeras

from hypernets.searchers.mcts_searcher import *
from hypernets.frameworks.keras.layers import *
from hypernets.frameworks.keras.hyper_keras import HyperKeras
from hypernets.core.callbacks import SummaryCallback
def get_space():
    space = HyperSpace()
    with space.as_default():
        in1 = Input(shape=(10,))
        dense1 = Dense(10, activation=Choice(['relu', 'tanh', None]), use_bias=Bool())(in1)
        bn1 = BatchNormalization()(dense1)
        dropout1 = Dropout(Choice([0.3, 0.4, 0.5]))(bn1)
        Dense(2, activation='softmax', use_bias=True)(dropout1)
    return space

mcts = MCTSSearcher(get_space, max_node_space=4)
hk = HyperKeras(mcts, optimizer='adam', loss='sparse_categorical_crossentropy', metrics=['accuracy'],
                callbacks=[SummaryCallback()])

x = np.random.randint(0, 10000, size=(100, 10))
y = np.random.randint(0, 2, size=(100), dtype='int')

hk.search(x, y, x, y, max_trails=10)
assert hk.get_best_trail()

HyperGBM

from hypernets.frameworks.ml.hyper_gbm import HyperGBM
from hypernets.frameworks.ml.common_ops import get_space_num_cat_pipeline_complex
from hypernets.searchers.random_searcher import RandomSearcher
from hypernets.core.callbacks import *
from hypernets.core.searcher import OptimizeDirection
from hypernets.frameworks.ml.datasets import dsutils
from sklearn.model_selection import train_test_split

rs = RandomSearcher(get_space_num_cat_pipeline_complex, optimize_direction=OptimizeDirection.Maximize)
hk = HyperGBM(rs, task='classification', reward_metric='accuracy',
              callbacks=[SummaryCallback(), FileLoggingCallback(rs)])

df = dsutils.load_bank()
df.drop(['id'], axis=1, inplace=True)
X_train, X_test = train_test_split(df.head(1000), test_size=0.2, random_state=42)
y_train = X_train.pop('y')
y_test = X_test.pop('y')

hk.search(X_train, y_train, X_test, y_test, max_trails=30)
best_trial = hk.get_best_trail()
assert best_trial

estimator = hk.final_train(best_trial.space_sample, X_train, y_train)
score = estimator.predict(X_test)
result = estimator.evaluate(X_test, y_test)

HyperDT (Hyperparameter Tuning & NAS for DeepTables)

Install DeepTables

pip install deeptables
from deeptables.models.hyper_dt import mini_dt_space, HyperDT
from hypernets.searchers.random_searcher import RandomSearcher
from hypernets.core.searcher import OptimizeDirection
from hypernets.core.callbacks import SummaryCallback, FileLoggingCallback
from hypernets.searchers.mcts_searcher import MCTSSearcher
from hypernets.searchers.evolution_searcher import EvolutionSearcher
from hypernets.core.trial import DiskTrailStore
from hypernets.frameworks.ml.datasets import dsutils
from sklearn.model_selection import train_test_split

disk_trail_store = DiskTrailStore('~/trail_store')

# searcher = MCTSSearcher(mini_dt_space, max_node_space=0,optimize_direction=OptimizeDirection.Maximize)
# searcher = RandomSearcher(mini_dt_space, optimize_direction=OptimizeDirection.Maximize)
searcher = EvolutionSearcher(mini_dt_space, 200, 100, regularized=True, candidates_size=30,
                             optimize_direction=OptimizeDirection.Maximize)

hdt = HyperDT(searcher, callbacks=[SummaryCallback(), FileLoggingCallback(searcher)], reward_metric='AUC',
              earlystopping_patience=1, )

df = dsutils.load_adult()
df_train, df_test = train_test_split(df, test_size=0.2, random_state=42)
X = df_train
y = df_train.pop(14)
y_test = df_test.pop(14)

hdt.search(df_train, y, df_test, y_test, max_trails=100, batch_size=256, epochs=10, verbose=1, )
best_trial = hdt.get_best_trail()
assert best_trial

DataCanvas

Hypernets is an open source project created by DataCanvas.

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

hypernets-0.1.6.tar.gz (63.5 kB view hashes)

Uploaded Source

Built Distribution

hypernets-0.1.6-py3-none-any.whl (95.2 kB view hashes)

Uploaded Python 3

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page