v0.2.6 (#8)

* TopBottomR similarity
* Update centrality tool
* Update ConceptTypicality for new version of fcapsy
* Centrality html export
* Black formatter

fcapsy_experiments/__init__.py

@@ -1,7 +1,7 @@
 from .correlation_table import CorrelationTable
 from .correlations_box_plot import correlations_boxplots
 
-__version__ = "0.2.5"
+__version__ = "0.2.6"
 __author__ = "Tomáš Mikula"
 __email__ = "mail@tomasmikula.cz"
-__license__ = "MIT license"
+__license__ = "MIT license"

fcapsy_experiments/_styles.py

@@ -24,10 +24,20 @@
 
 css_corr = [
     {"selector": "td, th", "props": [("text-align", "center")]},
-    # {"selector": "thead th", "props": [("text-align", "center !important")]},
+    {"selector": "thead th", "props": [("height", "50px"), ("text-align", "left")]},
     {"selector": "tbody th", "props": [("text-align", "right")]},
     {
         "selector": "th.col_heading",
-        "props": [("writing-mode", "vertical-rl"), ("transform", "rotateZ(-90deg)")],
+        "props": [
+            ("writing-mode", "vertical-rl"),
+            ("transform", "rotateZ(-90deg)"),
+            ("text-align", "left"),
+        ],
     },
 ]
+
+css_centrality = [
+    {"selector": "td, th", "props": [("text-align", "center")]},
+    {"selector": "thead th", "props": [("text-align", "center")]},
+    {"selector": "tbody th", "props": [("text-align", "right")]},
+]

fcapsy_experiments/centrality/__init__.py

@@ -1 +1 @@
-from .centrality import Centrality
+from .centrality import Centrality

fcapsy_experiments/centrality/centrality.py

@@ -1,7 +1,7 @@
 import pandas as pd
 
 from fcapsy.centrality import centrality
-from fcapsy_experiments._styles import css, css_corr
+from fcapsy_experiments._styles import css, css_centrality
 
 
 class Centrality:
@@ -95,6 +95,6 @@ def to_html(self, include_core_flag: bool = False, quantile: float = 0.75) -> st
 
         final_table = final_table.style.format(precision=3)
         final_table.background_gradient(cmap="RdYlGn")
-        final_table.set_table_styles(css + css_corr)
+        final_table.set_table_styles(css + css_centrality)
 
         return final_table.to_html()

fcapsy_experiments/mca/__init__.py

@@ -1 +1 @@
-from .mca_concept import MCAConcept
+from .mca_concept import MCAConcept

fcapsy_experiments/mca/mca_concept.py

@@ -57,22 +57,22 @@ def to_plotly(self) -> "go.Figure":
         data = self.concept_df_transformed.copy()
 
         hover_data = {
-                "x": False,
-                "y": False,
-            }
+            "x": False,
+            "y": False,
+        }
 
         if self._color_by is not None:
             color_name = self._color_by[0]
             data[color_name] = self._color_by[1]
             hover_data[color_name] = ":.2f"
         else:
             color_name = None
-            
+
         if self.n_components == 3:
             data = data.rename(columns={0: "x", 1: "y", 2: "z"})
 
             hover_data["z"] = False
-            
+
             fig = px.scatter_3d(
                 data,
                 x="x",
@@ -107,3 +107,23 @@ def to_plotly(self) -> "go.Figure":
             )
 
         return fig
+
+    def to_plotly_html(
+        self, default_width: int = 700, default_height: int = 390
+    ) -> str:
+        """Generates html version of plotly graph
+
+        Args:
+            default_width (int, optional): default graph width. Defaults to 700.
+            default_height (int, optional): default graph height. Defaults to 390.
+
+        Returns:
+            str: graph html
+        """
+        return self.to_plotly().to_html(
+            full_html=False,
+            include_plotlyjs="cdn",
+            include_mathjax="cdn",
+            default_width=default_width,
+            default_height=default_height,
+        )

fcapsy_experiments/typicality/__init__.py

@@ -1,2 +1,3 @@
 from .concept_typicality import ConceptTypicality
 from .top_r_similarity import TopRSimilarity
+from .top_bottom_r_similarity import TopBottomRSimilarity

fcapsy_experiments/typicality/concept_typicality.py

@@ -37,7 +37,11 @@ def __init__(
                 "typ_avg": {
                     # must be callable
                     "func": typicality_avg,
-                    "args": {"J": [jaccard], "SMC": [smc], "R": [russell_rao]},
+                    "args": {
+                        "J": {"similarity": jaccard},
+                        "SMC": {"similarity": smc},
+                        "R": {"similarity": russell_rao},
+                    },
                 }
             }
 
@@ -82,10 +86,10 @@ def _init(self, concept, count, typicality_functions, extra_columns):
             for typicality in typicality_functions.values():
                 function = typicality["func"]
                 args = typicality["args"].values()
+
                 if args:
-                    row = []
-                    for name, arg in typicality["args"].items():
-                        row.append(function(item, concept, *arg))
+                    for _, arg in typicality["args"].items():
+                        row.append(function(item, concept, **arg))
                 else:
                     row.append(function(item, concept))
 

fcapsy_experiments/typicality/top_bottom_r_similarity.py

@@ -0,0 +1,49 @@
+import typing
+
+import pandas as pd
+
+from itertools import combinations
+
+from .top_r_similarity import TopRSimilarity
+
+
+class TopBottomRSimilarity(TopRSimilarity):
+    @staticmethod
+    def _init(inst, to_columns):
+        r_range = range(1, len(inst._source.index))
+
+        results = []
+
+        # filtered_columns = filter(
+        #     lambda c: c not in ignore_columns, inst._source.columns
+        # )
+
+        columns_tuples, labels = inst._get_column_orders(
+            inst._source,
+            [(x, y) for x in inst._source.columns for y in to_columns if x != y],
+        )
+
+        for (column1_order, column2_order), label in zip(columns_tuples, labels):
+            column1_order_reversed = column1_order[::-1]
+            column2_order_reversed = column2_order[::-1]
+
+            for r in r_range:
+                top_r = inst._top_r_similarity(
+                    inst._context, inst._similarity, column1_order, column2_order, r
+                )
+                bottom_r = inst._top_r_similarity(
+                    inst._context,
+                    inst._similarity,
+                    column1_order_reversed,
+                    column2_order_reversed,
+                    r,
+                )
+                results.append(
+                    [
+                        r,
+                        top_r * bottom_r,
+                        label,
+                    ]
+                )
+
+        return pd.DataFrame(results, columns=["r", "top_bottom_r_similarity", "label"])

fcapsy_experiments/typicality/top_r_similarity.py

@@ -14,26 +14,26 @@ def __init__(
         source: "pd.DataFrame",
         context: "concepts.Context",
         similarity: typing.Callable = jaccard,
-        ignore_columns: typing.List[str] = None,
+        to_columns: typing.List[str] = None,
     ) -> None:
         """Calculates TopR similarities for given dataframe.
 
         Args:
             source (pd.DataFrame): source data (usually objects ordered by multiple metrics)
             context (concepts.Context): source formal context
             similarity (typing.Callable, optional): similarity which should be used. Defaults to jaccard.
-            ignore_columns (typing.List[str], optional): which columns to ignore from source dataframe. Defaults to None.
+            to_columns (typing.List[str], optional): which columns to compare every other from source dataframe. Defaults to None (means all).
         """
-        if ignore_columns is None:
-            ignore_columns = []
+        if to_columns is None:
+            to_columns = source.columns
 
         self._similarity = similarity
         self._source = source
         self._context = context
-        self.df = self._init(self, ignore_columns)
+        self.df = self._init(self, to_columns)
 
     @staticmethod
-    def _k_values_or_until_differs(iterator, r):
+    def _r_values_or_until_differs(iterator, r):
         max_idx = len(iterator) - 1
 
         if r == 0:
@@ -47,7 +47,26 @@ def _k_values_or_until_differs(iterator, r):
                 break
 
     @staticmethod
-    def _top_r_similarity(inst, metric_1_order, metric_2_order, r):
+    def _get_column_orders(df, tuples):
+        results = []
+        labels = []
+
+        for column1, column2 in tuples:
+            column1_order = list(
+                df.sort_values(column1, ascending=False, kind="mergesort").index
+            )
+
+            column2_order = list(
+                df.sort_values(column2, ascending=False, kind="mergesort").index
+            )
+
+            results.append([column1_order, column2_order])
+            labels.append(f"{column1}-{column2}")
+
+        return results, labels
+
+    @staticmethod
+    def _top_r_similarity(context, similarity, metric_1_order, metric_2_order, r):
         def _get_vectors(context, items):
             try:
                 label_domain = context._extents
@@ -62,56 +81,50 @@ def _get_vectors(context, items):
             return vectors
 
         vectors_1 = _get_vectors(
-            inst._context, list(inst._k_values_or_until_differs(metric_1_order, r))
+            context, list(TopRSimilarity._r_values_or_until_differs(metric_1_order, r))
         )
         vectors_2 = _get_vectors(
-            inst._context, list(inst._k_values_or_until_differs(metric_2_order, r))
+            context, list(TopRSimilarity._r_values_or_until_differs(metric_2_order, r))
         )
 
-        i1 = mean(
-            (max((inst._similarity(b1, b2) for b2 in vectors_1)) for b1 in vectors_2)
-        )
+        i1 = mean((max((similarity(b1, b2) for b2 in vectors_1)) for b1 in vectors_2))
 
-        i2 = mean(
-            (max((inst._similarity(b1, b2) for b2 in vectors_2)) for b1 in vectors_1)
-        )
+        i2 = mean((max((similarity(b1, b2) for b2 in vectors_2)) for b1 in vectors_1))
 
         return min(i1, i2)
 
     @staticmethod
-    def _init(inst, ignore_columns):
+    def _init(inst, to_columns):
         r_range = range(1, len(inst._source.index))
 
         results = []
 
-        filtered_columns = filter(
-            lambda c: c not in ignore_columns, inst._source.columns
-        )
-
-        for column1, column2 in combinations(filtered_columns, 2):
-            column1_order = list(
-                inst._source.sort_values(
-                    column1, ascending=False, kind="mergesort"
-                ).index
-            )
-            column2_order = list(
-                inst._source.sort_values(
-                    column2, ascending=False, kind="mergesort"
-                ).index
-            )
+        # filtered_columns = filter(
+        #     lambda c: c not in ignore_columns, inst._source.columns
+        # )
 
-            label = f"{column1}-{column2}"
+        columns_tuples, labels = inst._get_column_orders(
+            inst._source,
+            [(x, y) for x in inst._source.columns for y in to_columns if x != y],
+        )
 
+        for (column1_order, column2_order), label in zip(columns_tuples, labels):
             for r in r_range:
                 results.append(
                     [
                         r,
-                        inst._top_r_similarity(inst, column1_order, column2_order, r),
+                        inst._top_r_similarity(
+                            inst._context,
+                            inst._similarity,
+                            column1_order,
+                            column2_order,
+                            r,
+                        ),
                         label,
                     ]
                 )
 
-        return pd.DataFrame(results, columns=["r", "top_k_similarity", "label"])
+        return pd.DataFrame(results, columns=["r", "top_r_similarity", "label"])
 
     def to_plotly(self) -> "go.Figure":
         """Generates plotly figure.
@@ -122,7 +135,7 @@ def to_plotly(self) -> "go.Figure":
         fig = px.line(
             self.df,
             x="r",
-            y="top_k_similarity",
+            y=self.df.columns[1],
             color="label",
             labels={"label": "Legend"},
             line_dash="label",
@@ -148,6 +161,7 @@ def to_plotly(self) -> "go.Figure":
                 title="S",
                 mirror=True,
                 ticks="inside",
+                range=[0, 1],
                 showline=True,
                 linecolor="black",
                 linewidth=1,

setup.py

@@ -3,7 +3,7 @@
 
 setup(
     name="fcapsy-experiments",
-    version="0.2.4",
+    version="0.2.5",
     author="Tomáš Mikula",
     author_email="mail@tomasmikula.cz",
     description="Package of experiments for fcapsy library.",