takeshimeonerespect commented 4 years ago

Marcからtime series dataからpoint cloudを生成するソースコード等について、
以下の修正を依頼された。

[x] 名称の変更
[x] codeの追記、修正
[x] test

takeshimeonerespect commented 4 years ago

Docker container環境で動作確認、100%で終了することを確認済

takeshimeonerespect commented 4 years ago

Marcにreply済

takeshimeonerespect commented 4 years ago

2020-2-12

1次元配列、多次元配列の時系列データ処理の対応
- 1次元配列、多次元配列データの処理を分割
  - 1次元配列: __call__で実行
  - 多次元配列: transformで実行 __call__にif文による条件分岐を使用

class TimeDelayEmbedding:
    """Attractor transformation class for a single time-series data.

    Embeds time-series data in the R^d according to Takens' Embedding Theorem
    and obtains the coordinates of each point.

    Parameters
    ----------
    dim : int, optional (default=3)
        `d` of R^d to be embedded.

    delay : int, optional (default=1)
        Time-Delay embedding.

    skip : int, optional (default=1)
        How often to skip embedded points.

    """
    def __init__(self, dim=3, delay=1, skip=1):
        self._dim = dim
        self._delay = delay
        self._skip = skip

    def __call__(self, *args, **kwargs):
        ndts = np.array(ts)
        if ndts.ndim == 1:
            return self._transform(*args, **kwargs)
        else:
            return self.transform(*args, **kwargs)

    def fit(self, ts, y=None):
        return self

    def _transform(self, ts):
        """Guts of transform method.

        Parameters
        ----------
        ts : ndarray, shape=[n_features]
            A single time-series data.

        Returns
        -------
        attractor : ndarray, shape=[n_points, d]
            `n_points` is the number of points in the R^d.

        """
        ndts = np.array(ts)
        return ndts[
            np.add.outer(
                np.arange(0, len(ndts)-self._delay*(self._dim-1), self._skip),
                np.arange(0, self._dim*self._delay, self._delay))
        ]

    def transform(self, ts):
        """Transform method."""
        ndts = np.array(ts)
        return np.apply_along_axis(self._transform, 1, ndts)

takeshimeonerespect commented 4 years ago

2020-2-13

1次元配列、多次元配列の時系列データ処理の対応
- 1次元配列、多次元配列データ処理方法の見直し
  __call__でif文による条件分岐としてしまい、1次元配列を__call__で
  単独処理する形となっていないため。
  - 1次元配列: __call__で実行 if文による条件分岐を使用
  - 多次元配列: transformで実行 if文による条件分岐を使用
__init__.py追加

class TimeDelayEmbedding:
    """Point cloud transformation class.
    Embeds time-series data in the R^d according to Takens' Embedding Theorem
    and obtains the coordinates of each point.
    Parameters
    ----------
    dim : int, optional (default=3)
        `d` of R^d to be embedded.
    delay : int, optional (default=1)
        Time-Delay embedding.
    skip : int, optional (default=1)
        How often to skip embedded points.
        Given delay=3 and skip=2, an point cloud which is obtained by embedding 
        a single time-series data into R^3 is as follows.

    .. code-block:: none

    time-series = [1, 2, 3, 4, 5, 6, 7, 8, 9, 10]
    point clouds = [[1, 4, 7], 
                    [3, 6, 9]]

    """
    def __init__(self, dim=3, delay=1, skip=1):
        self._dim = dim
        self._delay = delay
        self._skip = skip

    def __call__(self, ts):
        """Transform method for single time-series data.
        Parameters
        ----------
        ts : list[float]
            A single time-series data.
        Returns
        -------
        point clouds : list[list[float, float, float]]
            Makes point cloud every a single time-series data.
        Raises
        -------
        TypeError
            If the parameter's type does not match the desired type.
        """
        ndts = np.array(ts)
        if ndts.ndim == 1:
            return self._transform(ndts)
        else:
            raise TypeError("Expects 1-dimensional array.")

    def fit(self, ts, y=None):
        return self

    def _transform(self, ts):
        """Guts of transform method."""
        return ts[
            np.add.outer(
                np.arange(0, len(ts)-self._delay*(self._dim-1), self._skip),
                np.arange(0, self._dim*self._delay, self._delay))
        ]

    def transform(self, ts):
        """Transform method for multiple time-series data.
        Parameters
        ----------
        ts : list[list[float]]
            Multiple time-series data.
        Attributes
        ----------
        ndts : 
             The ndts means that all time series need to have exactly 
             the same size. 
        Returns
        -------
        point clouds : list[list[list[float, float, float]]]
            Makes point cloud every a single time-series data.
        Raises
        -------
        TypeError
            If the parameter's type does not match the desired type.
        """
        ndts = np.array(ts)
        if ndts.ndim == 2:
            return np.apply_along_axis(self._transform, 1, ndts)
        else:
            raise TypeError("Expects 2-dimensional array.")

takeshimeonerespect commented 4 years ago

2020-2-14

test code(test_time_delay.py)の修正
- 変換の型をlistからnumpy arrayに変更したことにより、build時にerrorが発生
- ValueError: The truth value of an array with more than one element is ambiguous. Use a.any() or a.all() numpy
```
from gudhi.point_cloud.timedelay import TimeDelayEmbedding
```

def test_normal():

Sample array

ts = [1, 2, 3, 4, 5, 6, 7, 8, 9, 10]
# Normal case.
prep = TimeDelayEmbedding()
attractor = prep(ts)
assert (attractor[0] == [1, 2, 3])
assert (attractor[1] == [2, 3, 4])
assert (attractor[2] == [3, 4, 5])
assert (attractor[3] == [4, 5, 6])
assert (attractor[4] == [5, 6, 7])
assert (attractor[5] == [6, 7, 8])
assert (attractor[6] == [7, 8, 9])
assert (attractor[7] == [8, 9, 10])
# Delay = 3
prep = TimeDelayEmbedding(delay=3)
attractor = prep(ts)
assert (attractor[0] == [1, 4, 7])
assert (attractor[1] == [2, 5, 8])
assert (attractor[2] == [3, 6, 9])
assert (attractor[3] == [4, 7, 10])
# Skip = 3
prep = TimeDelayEmbedding(skip=3)
attractor = prep(ts)
assert (attractor[0] == [1, 2, 3])
assert (attractor[1] == [4, 5, 6])
assert (attractor[2] == [7, 8, 9])
# Delay = 2 / Skip = 2
prep = TimeDelayEmbedding(delay=2, skip=2)
attractor = prep(ts)
assert (attractor[0] == [1, 3, 5])
assert (attractor[1] == [3, 5, 7])
assert (attractor[2] == [5, 7, 9])

- np.arrayを追記
  - ValueError: The truth value of an array with more than one element is ambiguous. 
Use a.any() or a.all() numpy  
  - `np.array`を追記しても、errorは解消されず

from gudhi.point_cloud.timedelay import TimeDelayEmbedding import numpy as np

def test_normal():

Sample array

ts = [1, 2, 3, 4, 5, 6, 7, 8, 9, 10]
# Normal case.
prep = TimeDelayEmbedding()
attractor = prep(ts)
assert (attractor[0] == np.array([1, 2, 3]))
assert (attractor[1] == np.array([2, 3, 4]))
assert (attractor[2] == np.array([3, 4, 5]))
assert (attractor[3] == np.array([4, 5, 6]))
assert (attractor[4] == np.array([5, 6, 7]))
assert (attractor[5] == np.array([6, 7, 8]))
assert (attractor[6] == np.array([7, 8, 9]))
assert (attractor[7] == np.array([8, 9, 10]))
# Delay = 3
prep = TimeDelayEmbedding(delay=3)
attractor = prep(ts)
assert (attractor[0] == np.array([1, 4, 7]))
assert (attractor[1] == np.array([2, 5, 8]))
assert (attractor[2] == np.array([3, 6, 9]))
assert (attractor[3] == np.array([4, 7, 10]))
# Skip = 3
prep = TimeDelayEmbedding(skip=3)
attractor = prep(ts)
assert (attractor[0] == np.array([1, 2, 3]))
assert (attractor[1] == np.array([4, 5, 6]))
assert (attractor[2] == np.array([7, 8, 9]))
# Delay = 2 / Skip = 2
prep = TimeDelayEmbedding(delay=2, skip=2)
attractor = prep(ts)
assert (attractor[0] == np.array([1, 3, 5]))
assert (attractor[1] == np.array([3, 5, 7]))
assert (attractor[2] == np.array([5, 7, 9]))


- `.all()`を追記することによりerror解消

from gudhi.point_cloud.timedelay import TimeDelayEmbedding import numpy as np

def test_normal():

Sample array

ts = [1, 2, 3, 4, 5, 6, 7, 8, 9, 10]
# Normal case.
prep = TimeDelayEmbedding()
attractor = prep(ts)
assert (attractor[0] == np.array([1, 2, 3])).all()
assert (attractor[1] == np.array([2, 3, 4])).all()
assert (attractor[2] == np.array([3, 4, 5])).all()
assert (attractor[3] == np.array([4, 5, 6])).all()
assert (attractor[4] == np.array([5, 6, 7])).all()
assert (attractor[5] == np.array([6, 7, 8])).all()
assert (attractor[6] == np.array([7, 8, 9])).all()
assert (attractor[7] == np.array([8, 9, 10])).all()
# Delay = 3
prep = TimeDelayEmbedding(delay=3)
attractor = prep(ts)
assert (attractor[0] == np.array([1, 4, 7])).all()
assert (attractor[1] == np.array([2, 5, 8])).all()
assert (attractor[2] == np.array([3, 6, 9])).all()
assert (attractor[3] == np.array([4, 7, 10])).all()
# Skip = 3
prep = TimeDelayEmbedding(skip=3)
attractor = prep(ts)
assert (attractor[0] == np.array([1, 2, 3])).all()
assert (attractor[1] == np.array([4, 5, 6])).all()
assert (attractor[2] == np.array([7, 8, 9])).all()
# Delay = 2 / Skip = 2
prep = TimeDelayEmbedding(delay=2, skip=2)
attractor = prep(ts)
assert (attractor[0] == np.array([1, 3, 5])).all()
assert (attractor[1] == np.array([3, 5, 7])).all()
assert (attractor[2] == np.array([5, 7, 9])).all()

takeshimeonerespect commented 4 years ago

2020-2-19

dimに代わる新たな名称の提案
1. replicate
2. recursion
3. iteration
4. rebuild
5. recast
point cloudsの返却値(return)の修正
- old: point clouds : list[list[list[float, float, float]]]
- now: point clouds : list of n x 3 numpy arrays
多次元配列データのTimeDelayEmbedding時のデータの結合処理方法の検討
- np.hstackを使うとできそう
- リスト内包表記を使えばできそう [式 for 任意の変数名 in イテラブルオブジェクト] リストやタプルなどのイテラブルオブジェクトの各要素を任意の変数名で取り出し
  式で評価、その結果が要素となる新たなリストが返却される.

implementation

ndts = np.array(ts)
        if ndts.ndim == 2:
            join_array = np.apply_along_axis(self._transform, 1, ndts)
            return np.hstack([i for i in join_array])
        else:
            raise TypeError("Expects 2-dimensional array.")

takeshimeonerespect commented 4 years ago

2020-2-25

dimは、結局、名称変更なし

takeshimeonerespect commented 4 years ago

2020-2-26

Travis CIでError発生
- ドキュメント作成
- .travis.yml

env:

4. Only doxygen documentation

- CMAKE_EXAMPLE='OFF' CMAKE_TEST='OFF' CMAKE_UTILITIES='OFF' CMAKE_PYTHON='OFF' MAKE_TARGET='doxygen' CTEST_COMMAND='echo No tests for doxygen target'


doxygen: プログラムのコメントに書かれた情報をもとにドキュメントを作成するツール

失敗するテストは、C ++のドキュメント作成に関するもの
C++のドキュメントは、ブランチが触れないもの
すでにマスターに影響
これはmacosxにhomebrewを使用してdoxygenをインストールする際の問題
travisまたはhomebrewが何かを更新した後に表示される
Vincentが他のブランチで修正しようとしており、ブランチをブロックする必要はない

takeshimeonerespect commented 4 years ago

Close the issue because it was merged on March 16.

takeshimeonerespect / take_shimeone

GUDHIへのpull request後の対応 #2

2020-2-12

2020-2-13

2020-2-14

Sample array

Sample array

Sample array

2020-2-19

2020-2-25

2020-2-26

4. Only doxygen documentation