hyperparameter 0.3.0

A hyper-parameter library for researchers, data scientists and machine learning engineers.

HyperParameter

A hyper-parameter library for researchers, data scientists and machine learning engineers.

Quick Start

Object-Style API:

>>> from hyperparameter import HyperParameter

>>> hp = HyperParameter(a=1, b={'c': 2})
>>> hp.a == 1
True
>>> hp.b.c == 2  # (nested parameter)
True

If we want safe access to undefined parameters with default values, we can use hp() instead of hp:

>>> hp = HyperParameter()
>>> hp().a.b.c.get_or_else(3) # (default value for undefined parameter)
3
>>> hp().a.b.c(3)             # (shortcut for `get_or_else`)
3

>>> hp().a.b.c = 4          # set value to param `a.b.c`
>>> hp().a.b.c(3)           # (default value is ignored) 
4

Scoped Parameter

>>> from hyperparameter import param_scope

# scoped parameter
>>> with param_scope(a=1) as ps: 
...     ps.a == 1
True

When nested, the parameter modifications are limited to the inner scope:

>>> with param_scope(a=1) as ps:
...     with param_scope(a=2) as ps2:
...         ps2.a == 2 # True, a=2 for inner scope
...     ps.a == 1      # True, a=1 for outer scope
True
True

The nested scope feature can be used to config the default behavior when used in functions:

#change function behavior with scoped parameter:
def dnn(input):
    # receive parameter using param_scope
    with param_scope() as ps:
        output = linear(inputs)
        return activation_fn(
                    output, 
                    activation=ps().activation("relu"))

# call function with default parameter
dnn()

# passing parameter using param_scope
with param_scope(activation="sigmoid"): 
    dnn()

Predefined Parameter

@auto_param #convert keyword arguments into hyper parameters
def model_train(X, y, learning_rate = 1.0, penalty = 'l1'):
    LR = LogisticRegression(C=1.0, 
        lr=local_param('learning_rate'), 
        penalty=local_param('penalty'))
    LR.fit(X, y)

# specify predefined parameter using `param_scope`
with param_scope('model_train.learning_rate=0.01'):
    model_train(X, y)

Examples

parameter tunning for researchers

This example shows how to use hyperparameter in your research projects, and make your experiments reproducible.

experiment tracing for data scientists

This example shows experiment management with hyperparameter, and tracing the results with mlflow.tracing.

Todo.

design-pattern for system engineers

Todo.