test_feature_space.py

# Copyright 2021 Thomas A. R. Purcell
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import shutil
import numpy as np

from sissopp import (
    FeatureNode,
    FeatureSpace,
    Inputs,
    Unit,
    initialize_values_arr,
)

import pathlib

parent = pathlib.Path(__file__).parent.absolute()


def test_feature_space():
    task_sizes_train = [90]
    task_sizes_test = [10]

    initialize_values_arr(task_sizes_train, task_sizes_test, 10, 2)

    inputs = Inputs()
    inputs.phi_0 = [
        FeatureNode(
            ff,
            f"feat_{ff}",
            np.random.random(task_sizes_train[0]) * 1e2 - 50,
            np.random.random(task_sizes_test[0]) * 1e2 - 50,
            Unit("s"),
        )
        for ff in range(10)
    ]

    inputs.prop_train = np.power(inputs.phi_0[0].value + inputs.phi_0[1].value, 2.0)

    inputs.allowed_ops = ["add", "sub", "mult", "sq", "cb", "sqrt", "cbrt"]
    inputs.task_sizes_train = task_sizes_train
    inputs.task_sizes_test = task_sizes_test
    inputs.calc_type = "regression"
    inputs.max_rung = 2
    inputs.n_sis_select = 10
    inputs.phi_out_file = "feature_space/phi.txt"

    try:
        inputs.n_rung_generate = 2
        feat_space = FeatureSpace(inputs)
        raise ValueError("FeatureSpace created with invalid parameters")
    except RuntimeError:
        inputs.n_rung_generate = 0
        pass

    try:
        inputs.n_rung_generate = 1
        inputs.n_rung_store = 2
        feat_space = FeatureSpace(inputs)
        raise ValueError("FeatureSpace created with invalid parameters")
    except RuntimeError:
        inputs.n_rung_generate = 0
        inputs.n_rung_store = 1
        pass

    try:
        inputs.allowed_ops = ["exp"]
        feat_space = FeatureSpace(inputs)
        raise ValueError(
            "FeatureSpace created when there is a rung with no features created"
        )
    except RuntimeError:
        inputs.allowed_ops = ["add", "sub", "mult", "sq", "cb", "sqrt", "cbrt"]
        pass

    feat_space = FeatureSpace(inputs)
    assert feat_space.phi0[0].expr == inputs.phi_0[0].expr

    feat_space.sis(inputs.prop_train)

    assert feat_space.phi_selected[0].postfix_expr == "1|0|add|sq"

    feat_space.sis(list(inputs.prop_train))
    shutil.rmtree("feature_space/")
    assert feat_space.phi_selected[0].postfix_expr == "1|0|add|sq"

    assert (
        np.abs(
            np.corrcoef(inputs.prop_train, inputs.phi_0[0].value)[0, 1] ** 2.0
            + feat_space.scores[0]
        )
        < 1e-10
    )
    assert feat_space.task_sizes_train[0] == 90
    assert feat_space.allowed_param_ops == []
    assert feat_space.allowed_ops == ["add", "sub", "mult", "sq", "cb", "sqrt", "cbrt"]
    assert feat_space.start_rung[0] == 0
    assert feat_space.start_rung[1] == 10
    assert feat_space.get_feature(0).expr == "feat_0"
    assert feat_space.phi_selected[1].d_mat_ind == 1

    try:
        feat_space.remove_feature(feat_space.phi_selected[0].feat_ind)
        raise ValueError("Removed selected feature.")
    except RuntimeError:
        pass

    test_expr = feat_space.get_feature(len(feat_space.phi) - 2).expr
    feat_space.remove_feature(len(feat_space.phi) - 2)
    assert feat_space.get_feature(len(feat_space.phi) - 2).expr != test_expr

    if feat_space.parameterized_feats_allowed:
        try:
            inputs.max_param_depth = 10
            feat_space = FeatureSpace(inputs)
            raise ValueError("FeatureSpace created with invalid parameters")
        except RuntimeError:
            inputs.max_param_depth = 0
            pass

    feat_space.output_phi()
    feat_space_2 = FeatureSpace(
        f"{parent}/feature_space/phi.txt",
        inputs.phi_0,
        inputs.prop_train,
        inputs.task_sizes_train,
        inputs.calc_type,
        inputs.n_sis_select,
        inputs.cross_cor_max,
    )

    # shutil.rmtree("feature_space/")
    assert feat_space.n_feat == feat_space_2.n_feat

    feat_space_2.sis(inputs.prop_train)
    assert (
        feat_space.phi_selected[0].postfix_expr
        == feat_space_2.phi_selected[0].postfix_expr
    )


if __name__ == "__main__":
    test_feature_space()