SimThresholdSplit.similarity_matrix() - Code Metrics - Inspection of "added docstring and license to vis/mol" - richlewis42/scikit-chem - Measure and Improve Code Quality continuously with Scrutinizer

Completed

Push — master ( 8ca4b8...1dc969 )

by Rich

created 2016-06-03 13:34 UTC

SimThresholdSplit.similarity_matrix() A

↳ Parent: SimThresholdSplit

Complexity

Conditions

Size

Total Lines

Duplication

Lines	0
Ratio	0 %

Importance

Changes	1
Bugs	0	Features	1

Metric	Value
cc	2
c	1
b	0
f	1
dl	0
loc	7
rs	9.4285

#! /usr/bin/env python
#
# Copyright (C) 2016 Rich Lewis <[email protected]>
# License: 3-clause BSD

"""
## skchem.cross_validation.similarity_threshold

Similarity threshold dataset partitioning functionality.
"""


import numpy as np
# .scrutinizer.yml
before_commands:
    - sudo pip install abc # Python2
    - sudo pip3 install abc # Python3
import pandas as pd
# .scrutinizer.yml
before_commands:
    - sudo pip install abc # Python2
    - sudo pip3 install abc # Python3

from scipy.spatial.distance import pdist, squareform, cdist
# .scrutinizer.yml
before_commands:
    - sudo pip install abc # Python2
    - sudo pip3 install abc # Python3
from scipy.sparse import dok_matrix, triu
# .scrutinizer.yml
before_commands:
    - sudo pip install abc # Python2
    - sudo pip3 install abc # Python3

from .. import descriptors


class SimThresholdSplit(object):
class SomeClass:
    def some_method(self):
        """Do x and return foo."""

    def __init__(self, inp, threshold=0.5, fper='morgan',

                 similarity_metric='jaccard', memory_optimized=False,
                 fingerprints=None, similarity_matrix=None):
        """ Threshold similarity split for chemical datasets.

        This class implements a splitting technique that will pool compounds
        with similarity above a theshold into the same splits.

        Machine learning techniques should be able to extrapolate outside of a
        molecular series, or scaffold, however random splits will result in some
        'easy' test sets that are either *identical* or in the same molecular
        series or share a significant scaffold with training set compounds.

        This splitting technique reduces or eliminates (depending on the
        threshold set) this effect, making the problem harder.

        Args:
            inp (scipy.sparse.dok, pd.Series or pd.DataFrame):
                Either:
                - a series of skchem.Mols
                - dataframe of precalculated fingerprints

            n_splits (int):
                The number of splits to give.  This will be overridden if ratio
                is passed.

            ratio (list[floats]):
                Split ratios to use.

            threshold (float):
                The similarity threshold, above which, compounds will all be
                assigned to the same split.

            fper (str or skchem.Fingerprinter):
                The fingerprinting technique to use to generate the similarity
                matrix.

            fingerprints (bool):
                Whether percalculated fingerprints were passed directly.

            similarity_matrix (scipy.sparse.dok):
                A precalculated similarity matrix.

        Notes:
            The splits will not always be exactly the size requested, due to the
            constraint and requirement to maintain random shuffling.
        """

        if isinstance(fper, str):
            fper = descriptors.get(fper)

        self.fper = fper
        fps = inp if fingerprints else self.fper.transform(inp)

        self.n_instances = len(inp)

        self.threshold = threshold
        self.similarity_metric = similarity_metric
        self.memory_optimized = memory_optimized

        if not similarity_matrix:
            similarity_matrix = self.similarity_matrix(fps)

        self.clusters = pd.Series(self._cluster(similarity_matrix),
                                  index=fps.index,
                                  name='clusters')

    def _cluster_cumsum(self, shuffled=True):

        nums = self.clusters.value_counts()
        if shuffled:
            nums = nums.ix[np.random.permutation(nums.index)].cumsum()
        return nums

    def split(self, ratio):

        """ Return splits of the data with thresholded similarity according to a
        specified ratio.

        Args:
            ratio (tuple[ints]):
                the ratio to use.
        Returns:
            generator[pd.Series]:
                Generator of boolean split masks for the reqested splits.

        Example:
            st = SimThresholdSplit(ms, fper='morgan', similarity_metric='jaccard')
            train, valid, test = st.split(ratio=(70, 15, 15))
        """

        ratio = self.split_sizes(ratio)
        nums = self._cluster_cumsum()
        res = pd.Series(np.nan, index=nums.index, name='split')

        for i, _ in enumerate(ratio):
            lower = 0 if i == 0 else sum(ratio[:i])
            upper = len(ratio) if i == len(ratio) else sum(ratio[:i + 1])
            res[nums[(nums > lower) & (nums <= upper)].index] = i

        res = res.sort_index()
        res = self.clusters.to_frame().join(res, on='clusters')['split']
        return (res == i for i, _ in enumerate(ratio))

    def k_fold(self, n_folds):

        """ Returns k-fold cross-validated folds with thresholded similarity.

        Args:
            n_folds (int):
                The number of folds to provide.

        Returns:
            generator[(pd.Series, pd.Series)]:
                The splits in series.
        """

        folds = self.split((1,) * n_folds)
        return ((~fold, fold) for fold in folds)


    def split_sizes(self, ratio):
        """ Calculate the sizes of the splits """

        tot = sum(ratio)
        return [self.n_instances * rat / tot for rat in ratio]


    def similarity_matrix(self, fps):
        """ Calculate the similarity matrix for fingerprints. """

        if self.memory_optimized:
            return self._sim_low_mem(fps)
        else:
            return self._sim(fps)


    def _sim(self, fps):
        """ Fast but memory intensive implementation of similarity matrix
        calculation. """

        D = squareform(pdist(fps, self.similarity_metric))

        D = 1 - D # similarity is 1 - distance

        return triu(D >= self.threshold, k=1).todok()

    def _sim_low_mem(self, fps):
        """ Slow but memory efficient implementation of similarity matrix
        calculation """

        S = dok_matrix(len(fps), len(fps))

        for i, fp in enumerate(fps):

            D = cdist(fp[np.newaxis, :], fps[i + 1:], self.similarity_metric)

            D = 1 - D

            S[i, i + 1:] = dok_matrix(D >= threshold)

        return S

    def _cluster(self, S):

        """ Assign instances to clusters. """

        pairs = sorted(S.keys(), key=lambda x: x[0]) # sort pairs by first index
        clustered = np.arange(self.n_instances)

        for i, j in pairs:
            clustered[j] = clustered[i]

        return clustered


Push — master ( 8ca4b8...1dc969 )

SimThresholdSplit.similarity_matrix() A

Complexity

Size

Duplication

Importance

1. Missing Dependencies

2. Missing init.py files

1. Missing Dependencies

2. Missing init.py files

1. Missing Dependencies

2. Missing init.py files

1. Missing Dependencies

2. Missing init.py files

1			#! /usr/bin/env python
2			#
3			# Copyright (C) 2016 Rich Lewis <[email protected]>
4			# License: 3-clause BSD
5
6			"""
7			## skchem.cross_validation.similarity_threshold
8
9			Similarity threshold dataset partitioning functionality.
10			"""
11
12
13			import numpy as np
			0 ignored issues – show Configuration introduced 2016-06-03 13:36 UTC by Report Bug Copy Issue Report The import `numpy` could not be resolved. This can be caused by one of the following: 1. Missing Dependencies This error could indicate a configuration issue of Pylint. Make sure that your libraries are available by adding the necessary commands. # .scrutinizer.yml before_commands: - sudo pip install abc # Python2 - sudo pip3 install abc # Python3 Tip: We are currently not using virtualenv to run pylint, when installing your modules make sure to use the command for the correct version. 2. Missing __init__.py files This error could also result from missing `__init__.py` files in your module folders. Make sure that you place one file in each sub-folder. Loading history...
14			import pandas as pd
			0 ignored issues – show Configuration introduced 2016-06-03 13:36 UTC by Report Bug Copy Issue Report The import `pandas` could not be resolved. This can be caused by one of the following: 1. Missing Dependencies This error could indicate a configuration issue of Pylint. Make sure that your libraries are available by adding the necessary commands. # .scrutinizer.yml before_commands: - sudo pip install abc # Python2 - sudo pip3 install abc # Python3 Tip: We are currently not using virtualenv to run pylint, when installing your modules make sure to use the command for the correct version. 2. Missing __init__.py files This error could also result from missing `__init__.py` files in your module folders. Make sure that you place one file in each sub-folder. Loading history...
15
16			from scipy.spatial.distance import pdist, squareform, cdist
			0 ignored issues – show Configuration introduced 2016-06-03 13:36 UTC by Report Bug Copy Issue Report The import `scipy.spatial.distance` could not be resolved. This can be caused by one of the following: 1. Missing Dependencies This error could indicate a configuration issue of Pylint. Make sure that your libraries are available by adding the necessary commands. # .scrutinizer.yml before_commands: - sudo pip install abc # Python2 - sudo pip3 install abc # Python3 Tip: We are currently not using virtualenv to run pylint, when installing your modules make sure to use the command for the correct version. 2. Missing __init__.py files This error could also result from missing `__init__.py` files in your module folders. Make sure that you place one file in each sub-folder. Loading history...
17			from scipy.sparse import dok_matrix, triu
			0 ignored issues – show Configuration introduced 2016-06-03 13:36 UTC by Report Bug Copy Issue Report The import `scipy.sparse` could not be resolved. This can be caused by one of the following: 1. Missing Dependencies This error could indicate a configuration issue of Pylint. Make sure that your libraries are available by adding the necessary commands. # .scrutinizer.yml before_commands: - sudo pip install abc # Python2 - sudo pip3 install abc # Python3 Tip: We are currently not using virtualenv to run pylint, when installing your modules make sure to use the command for the correct version. 2. Missing __init__.py files This error could also result from missing `__init__.py` files in your module folders. Make sure that you place one file in each sub-folder. Loading history...
18
19			from .. import descriptors
20
21
22			class SimThresholdSplit(object):
			0 ignored issues – show Coding Style introduced 2016-06-03 13:36 UTC by Report Bug Copy Issue Report This class should have a docstring. The coding style of this project requires that you add a docstring to this code element. Below, you find an example for methods: class SomeClass: def some_method(self): """Do x and return foo.""" If you would like to know more about docstrings, we recommend to read PEP-257: Docstring Conventions. Loading history...
23
24			def __init__(self, inp, threshold=0.5, fper='morgan',
			0 ignored issues – show best-practice introduced 2016-06-03 13:36 UTC by Report Bug Copy Issue Report Too many arguments (8/5) Loading history...
25			similarity_metric='jaccard', memory_optimized=False,
26			fingerprints=None, similarity_matrix=None):
27			""" Threshold similarity split for chemical datasets.
28
29			This class implements a splitting technique that will pool compounds
30			with similarity above a theshold into the same splits.
31
32			Machine learning techniques should be able to extrapolate outside of a
33			molecular series, or scaffold, however random splits will result in some
34			'easy' test sets that are either identical or in the same molecular
35			series or share a significant scaffold with training set compounds.
36
37			This splitting technique reduces or eliminates (depending on the
38			threshold set) this effect, making the problem harder.
39
40			Args:
41			inp (scipy.sparse.dok, pd.Series or pd.DataFrame):
42			Either:
43			- a series of skchem.Mols
44			- dataframe of precalculated fingerprints
45
46			n_splits (int):
47			The number of splits to give. This will be overridden if ratio
48			is passed.
49
50			ratio (list[floats]):
51			Split ratios to use.
52
53			threshold (float):
54			The similarity threshold, above which, compounds will all be
55			assigned to the same split.
56
57			fper (str or skchem.Fingerprinter):
58			The fingerprinting technique to use to generate the similarity
59			matrix.
60
61			fingerprints (bool):
62			Whether percalculated fingerprints were passed directly.
63
64			similarity_matrix (scipy.sparse.dok):
65			A precalculated similarity matrix.
66
67			Notes:
68			The splits will not always be exactly the size requested, due to the
69			constraint and requirement to maintain random shuffling.
70			"""
71
72			if isinstance(fper, str):
73			fper = descriptors.get(fper)
74
75			self.fper = fper
76			fps = inp if fingerprints else self.fper.transform(inp)
77
78			self.n_instances = len(inp)
79
80			self.threshold = threshold
81			self.similarity_metric = similarity_metric
82			self.memory_optimized = memory_optimized
83
84			if not similarity_matrix:
85			similarity_matrix = self.similarity_matrix(fps)
86
87			self.clusters = pd.Series(self._cluster(similarity_matrix),
88			index=fps.index,
89			name='clusters')
90
91			def _cluster_cumsum(self, shuffled=True):
92
93			nums = self.clusters.value_counts()
94			if shuffled:
95			nums = nums.ix[np.random.permutation(nums.index)].cumsum()
96			return nums
97
98			def split(self, ratio):
99
100			""" Return splits of the data with thresholded similarity according to a
101			specified ratio.
102
103			Args:
104			ratio (tuple[ints]):
105			the ratio to use.
106			Returns:
107			generator[pd.Series]:
108			Generator of boolean split masks for the reqested splits.
109
110			Example:
111			st = SimThresholdSplit(ms, fper='morgan', similarity_metric='jaccard')
112			train, valid, test = st.split(ratio=(70, 15, 15))
113			"""
114
115			ratio = self.split_sizes(ratio)
116			nums = self._cluster_cumsum()
117			res = pd.Series(np.nan, index=nums.index, name='split')
118
119			for i, _ in enumerate(ratio):
120			lower = 0 if i == 0 else sum(ratio[:i])
121			upper = len(ratio) if i == len(ratio) else sum(ratio[:i + 1])
122			res[nums[(nums > lower) & (nums <= upper)].index] = i
123
124			res = res.sort_index()
125			res = self.clusters.to_frame().join(res, on='clusters')['split']
126			return (res == i for i, _ in enumerate(ratio))
127
128			def k_fold(self, n_folds):
129
130			""" Returns k-fold cross-validated folds with thresholded similarity.
131
132			Args:
133			n_folds (int):
134			The number of folds to provide.
135
136			Returns:
137			generator[(pd.Series, pd.Series)]:
138			The splits in series.
139			"""
140
141			folds = self.split((1,) * n_folds)
142			return ((~fold, fold) for fold in folds)
143
144
145			def split_sizes(self, ratio):
146			""" Calculate the sizes of the splits """
147
148			tot = sum(ratio)
149			return [self.n_instances * rat / tot for rat in ratio]
150
151
152			def similarity_matrix(self, fps):
153			""" Calculate the similarity matrix for fingerprints. """
154
155			if self.memory_optimized:
156			return self._sim_low_mem(fps)
157			else:
158			return self._sim(fps)
159
160
161			def _sim(self, fps):
162			""" Fast but memory intensive implementation of similarity matrix
163			calculation. """
164
165			D = squareform(pdist(fps, self.similarity_metric))
			0 ignored issues – show Coding Style Naming introduced 2016-06-03 13:36 UTC by Report Bug Copy Issue Report The name `D` does not conform to the variable naming conventions (`[a-z_][a-z0-9_]{2,30}$`). This check looks for invalid names for a range of different identifiers. You can set regular expressions to which the identifiers must conform if the defaults do not match your requirements. If your project includes a Pylint configuration file, the settings contained in that file take precedence. To find out more about Pylint, please refer to their site. Loading history...
166			D = 1 - D # similarity is 1 - distance
			0 ignored issues – show Coding Style Naming introduced 2016-06-03 13:36 UTC by Report Bug Copy Issue Report The name `D` does not conform to the variable naming conventions (`[a-z_][a-z0-9_]{2,30}$`). This check looks for invalid names for a range of different identifiers. You can set regular expressions to which the identifiers must conform if the defaults do not match your requirements. If your project includes a Pylint configuration file, the settings contained in that file take precedence. To find out more about Pylint, please refer to their site. Loading history...
167			return triu(D >= self.threshold, k=1).todok()
168
169			def _sim_low_mem(self, fps):
170			""" Slow but memory efficient implementation of similarity matrix
171			calculation """
172
173			S = dok_matrix(len(fps), len(fps))
			0 ignored issues – show Coding Style Naming introduced 2016-06-03 13:36 UTC by Report Bug Copy Issue Report The name `S` does not conform to the variable naming conventions (`[a-z_][a-z0-9_]{2,30}$`). This check looks for invalid names for a range of different identifiers. You can set regular expressions to which the identifiers must conform if the defaults do not match your requirements. If your project includes a Pylint configuration file, the settings contained in that file take precedence. To find out more about Pylint, please refer to their site. Loading history...
174			for i, fp in enumerate(fps):
			0 ignored issues – show Coding Style Naming introduced 2016-06-03 13:36 UTC by Report Bug Copy Issue Report The name `fp` does not conform to the variable naming conventions (`[a-z_][a-z0-9_]{2,30}$`). This check looks for invalid names for a range of different identifiers. You can set regular expressions to which the identifiers must conform if the defaults do not match your requirements. If your project includes a Pylint configuration file, the settings contained in that file take precedence. To find out more about Pylint, please refer to their site. Loading history...
175			D = cdist(fp[np.newaxis, :], fps[i + 1:], self.similarity_metric)
			0 ignored issues – show Coding Style Naming introduced 2016-06-03 13:36 UTC by Report Bug Copy Issue Report The name `D` does not conform to the variable naming conventions (`[a-z_][a-z0-9_]{2,30}$`). This check looks for invalid names for a range of different identifiers. You can set regular expressions to which the identifiers must conform if the defaults do not match your requirements. If your project includes a Pylint configuration file, the settings contained in that file take precedence. To find out more about Pylint, please refer to their site. Loading history...
176			D = 1 - D
			0 ignored issues – show Coding Style Naming introduced 2016-06-03 13:36 UTC by Report Bug Copy Issue Report The name `D` does not conform to the variable naming conventions (`[a-z_][a-z0-9_]{2,30}$`). This check looks for invalid names for a range of different identifiers. You can set regular expressions to which the identifiers must conform if the defaults do not match your requirements. If your project includes a Pylint configuration file, the settings contained in that file take precedence. To find out more about Pylint, please refer to their site. Loading history...
177			S[i, i + 1:] = dok_matrix(D >= threshold)
			0 ignored issues – show Comprehensibility Best Practice introduced 2016-06-03 13:36 UTC by Report Bug Copy Issue Report Undefined variable 'threshold' Loading history...
178			return S
179
180			def _cluster(self, S):
			0 ignored issues – show Coding Style Naming introduced 2016-06-03 13:36 UTC by Report Bug Copy Issue Report The name `S` does not conform to the argument naming conventions (`[a-z_][a-z0-9_]{2,30}$`). This check looks for invalid names for a range of different identifiers. You can set regular expressions to which the identifiers must conform if the defaults do not match your requirements. If your project includes a Pylint configuration file, the settings contained in that file take precedence. To find out more about Pylint, please refer to their site. Loading history...
181			""" Assign instances to clusters. """
182
183			pairs = sorted(S.keys(), key=lambda x: x[0]) # sort pairs by first index
184			clustered = np.arange(self.n_instances)
185
186			for i, j in pairs:
187			clustered[j] = clustered[i]
188
189			return clustered
190

richlewis42 / scikit-chem

Push — master ( 8ca4b8...1dc969 )

SimThresholdSplit.similarity_matrix() A

Complexity

Size

Duplication

Importance

1. Missing Dependencies

2. Missing __init__.py files

1. Missing Dependencies

2. Missing __init__.py files

1. Missing Dependencies

2. Missing __init__.py files

1. Missing Dependencies

2. Missing __init__.py files

Duplication Side-by-Side

Filter issues like

2. Missing init.py files

2. Missing init.py files

2. Missing init.py files

2. Missing init.py files