Rename m to k

README.rst

@@ -39,7 +39,7 @@ Create an undirected network with preferential attachment and homophily.
 .. code:: pycon
 
     >>> from netin.models import PAHModel
-    >>> m = PAHModel(n=200, m=2, f_m=0.2, h_MM=0.1, h_mm=0.9, seed=42)
+    >>> m = PAHModel(n=200, k=2, f_m=0.2, h_MM=0.1, h_mm=0.9, seed=42)
     >>> m.simulate()
 
 

examples/notebooks/inference.ipynb

@@ -43,7 +43,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 2,
+   "execution_count": null,
    "metadata": {
     "ExecuteTime": {
      "end_time": "2024-09-11T10:26:35.083384Z",
@@ -53,7 +53,7 @@
    "outputs": [],
    "source": [
     "OUTPUT_DIR = 'inference_results'\n",
-    "FIGSIZE = (6,3) "
+    "FIGSIZE = (6,3)"
    ]
   },
   {
@@ -65,7 +65,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 3,
+   "execution_count": null,
    "metadata": {
     "ExecuteTime": {
      "end_time": "2024-09-11T10:26:35.122852Z",
@@ -75,7 +75,7 @@
    "outputs": [],
    "source": [
     "N=1000\n",
-    "m=2\n",
+    "k=2\n",
     "f_m=0.5\n",
     "h_mm=0.2\n",
     "h_MM=0.2\n",
@@ -84,11 +84,11 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 4,
+   "execution_count": null,
    "metadata": {},
    "outputs": [],
    "source": [
-    "g = PAHModel(N=N, m=2, f_m=f_m, h_mm=h_mm, h_MM=h_MM, seed=seed)\n",
+    "g = PAHModel(N=N, k=2, f_m=f_m, h_mm=h_mm, h_MM=h_MM, seed=seed)\n",
     "g = g.simulate()"
    ]
   },
@@ -251,7 +251,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 13,
+   "execution_count": null,
    "metadata": {
     "ExecuteTime": {
      "end_time": "2024-09-11T10:27:23.555466Z",
@@ -271,7 +271,7 @@
     }
    ],
    "source": [
-    "j.plot_evidences(bayes_factors=True, \n",
+    "j.plot_evidences(bayes_factors=True,\n",
     "                 figsize=FIGSIZE, bboxx=1.4,\n",
     "                 output_dir=OUTPUT_DIR)"
    ]

examples/notebooks/ranking.ipynb

@@ -61,10 +61,10 @@
     "    return f\"{df.name}\\n{', '.join(g)}\"\n",
     "\n",
     "def create_pah(\n",
-    "        n: int, f_m: float, m: int,\n",
+    "        n: int, f_m: float, k: int,\n",
     "        h_MM: float, h_mm: float,\n",
     "        seed: Union[int, Any, None] = None):\n",
-    "    model = PAHModel(n=n, f_m=f_m, m=m, h_MM=h_MM, h_mm=h_mm, seed=seed)\n",
+    "    model = PAHModel(n=n, f_m=f_m, k=k, h_MM=h_MM, h_mm=h_mm, seed=seed)\n",
     "    graph = model.simulate()\n",
     "    df = stats.get_node_metadata_as_dataframe(\n",
     "        graph=graph, include_graph_metadata=False)\n",
@@ -103,7 +103,7 @@
    "source": [
     "n = 500\n",
     "f_m = 0.5\n",
-    "m = 2\n",
+    "k = 2\n",
     "d = 0.005\n",
     "plo_M = 2.0\n",
     "plo_m = 2.0\n",
@@ -130,7 +130,7 @@
     "h_mm = 0.5\n",
     "\n",
     "# pah\n",
-    "df = create_pah(n=n, f_m=f_m, m=m, h_MM=h_MM, h_mm=h_mm, seed=seed)\n",
+    "df = create_pah(n=n, f_m=f_m, k=k, h_MM=h_MM, h_mm=h_mm, seed=seed)\n",
     "node_metadata_balance_pah.append(df)\n",
     "\n",
     "#daph\n",
@@ -155,7 +155,7 @@
     "h_mm = 0.9\n",
     "\n",
     "# pah\n",
-    "df = create_pah(n=n, f_m=f_m, m=m, h_MM=h_MM, h_mm=h_mm, seed=seed)\n",
+    "df = create_pah(n=n, f_m=f_m, k=k, h_MM=h_MM, h_mm=h_mm, seed=seed)\n",
     "node_metadata_balance_pah.append(df)\n",
     "\n",
     "#daph\n",
@@ -180,7 +180,7 @@
     "h_mm = 0.1\n",
     "\n",
     "# pah\n",
-    "df = create_pah(n=n, f_m=f_m, m=m, h_MM=h_MM, h_mm=h_mm, seed=seed)\n",
+    "df = create_pah(n=n, f_m=f_m, k=k, h_MM=h_MM, h_mm=h_mm, seed=seed)\n",
     "node_metadata_balance_pah.append(df)\n",
     "\n",
     "#daph\n",
@@ -205,7 +205,7 @@
     "h_mm = 0.1\n",
     "\n",
     "# pah\n",
-    "df = create_pah(n=n, f_m=f_m, m=m, h_MM=h_MM, h_mm=h_mm, seed=seed)\n",
+    "df = create_pah(n=n, f_m=f_m, k=k, h_MM=h_MM, h_mm=h_mm, seed=seed)\n",
     "node_metadata_balance_pah.append(df)\n",
     "\n",
     "#daph\n",
@@ -230,7 +230,7 @@
     "h_mm = 0.9\n",
     "\n",
     "# pah\n",
-    "df = create_pah(n=n, f_m=f_m, m=m, h_MM=h_MM, h_mm=h_mm, seed=seed)\n",
+    "df = create_pah(n=n, f_m=f_m, k=k, h_MM=h_MM, h_mm=h_mm, seed=seed)\n",
     "node_metadata_balance_pah.append(df)\n",
     "\n",
     "#daph\n",
@@ -274,7 +274,7 @@
     "h_mm = 0.5\n",
     "\n",
     "# pah\n",
-    "df = create_pah(n=n, f_m=f_m, m=m, h_MM=h_MM, h_mm=h_mm, seed=seed)\n",
+    "df = create_pah(n=n, f_m=f_m, k=k, h_MM=h_MM, h_mm=h_mm, seed=seed)\n",
     "node_metadata_unbalance_pah.append(df)\n",
     "\n",
     "#daph\n",
@@ -299,7 +299,7 @@
     "h_mm = 0.9\n",
     "\n",
     "# pah\n",
-    "df = create_pah(n=n, f_m=f_m, m=m, h_MM=h_MM, h_mm=h_mm, seed=seed)\n",
+    "df = create_pah(n=n, f_m=f_m, k=k, h_MM=h_MM, h_mm=h_mm, seed=seed)\n",
     "node_metadata_unbalance_pah.append(df)\n",
     "\n",
     "#daph\n",
@@ -324,7 +324,7 @@
     "h_mm = 0.1\n",
     "\n",
     "# pah\n",
-    "df = create_pah(n=n, f_m=f_m, m=m, h_MM=h_MM, h_mm=h_mm, seed=seed)\n",
+    "df = create_pah(n=n, f_m=f_m, k=k, h_MM=h_MM, h_mm=h_mm, seed=seed)\n",
     "node_metadata_unbalance_pah.append(df)\n",
     "\n",
     "#daph\n",
@@ -349,7 +349,7 @@
     "h_mm = 0.1\n",
     "\n",
     "# pah\n",
-    "df = create_pah(n=n, f_m=f_m, m=m, h_MM=h_MM, h_mm=h_mm, seed=seed)\n",
+    "df = create_pah(n=n, f_m=f_m, k=k, h_MM=h_MM, h_mm=h_mm, seed=seed)\n",
     "node_metadata_unbalance_pah.append(df)\n",
     "\n",
     "#daph\n",
@@ -374,7 +374,7 @@
     "h_mm = 0.9\n",
     "\n",
     "# pah\n",
-    "df = create_pah(n=n, f_m=f_m, m=m, h_MM=h_MM, h_mm=h_mm, seed=seed)\n",
+    "df = create_pah(n=n, f_m=f_m, k=k, h_MM=h_MM, h_mm=h_mm, seed=seed)\n",
     "node_metadata_unbalance_pah.append(df)\n",
     "\n",
     "#daph\n",

examples/undirected/barabasi_albert.py

@@ -3,11 +3,11 @@
 
 def run():
     n = 200
-    m = 2
+    k = 2
     seed = 1234
 
     model = BarabasiAlbertModel(n=n,
-                                m=m,
+                                k=k,
                                 seed=seed)
     timer = SimulationTimer(model)
     model.simulate()

examples/undirected/homophily.py

@@ -3,13 +3,13 @@
 
 def run():
     n = 4000
-    m = 2
+    k = 2
     f_m = 0.1
     h_MM = 0.5
     h_mm = 0.5
     seed = 1234
 
-    model = HomophilyModel(n=n, m=m, f_m=f_m,
+    model = HomophilyModel(n=n, k=k, f_m=f_m,
                            h_mm=h_mm, h_MM=h_MM,
                            seed=seed)
     timer = SimulationTimer(model)

examples/undirected/pa.py

@@ -4,12 +4,12 @@
 
 def run():
     n = 4000
-    m = 2
+    k = 2
     f_m = 0.1
     seed = 1234
 
     model = PAModel(n=n,
-                    m=m,
+                    k=k,
                     f_m=f_m,
                     seed=seed)
     timer = SimulationTimer(model)

examples/undirected/pah.py

@@ -3,13 +3,13 @@
 
 def run():
     n = 4000
-    m = 2
+    k = 2
     f_m = 0.1
     h_MM = 0.5
     h_mm = 0.5
     seed = 1234
 
-    model = PAHModel(n=n, m=m, f_m=f_m,
+    model = PAHModel(n=n, k=k, f_m=f_m,
                      h_mm=h_mm, h_MM=h_MM,
                      seed=seed)
     timer = SimulationTimer(model)

examples/undirected/patch.py

@@ -12,7 +12,7 @@ def run():
     lfm_g = CompoundLFM.PAH
     seed = 1234
 
-    model = PATCHModel(n=n, m=k, f_m=f_m,
+    model = PATCHModel(n=n, k=k, f_m=f_m,
                        p_tc=p_tc,
                        lfm_local=lfm_l, lfm_global=lfm_g,
                        lfm_params={'h_mm': h_mm, 'h_MM': h_MM},

examples/undirected/testing.py

@@ -0,0 +1,121 @@
+from typing import Dict, Tuple
+from netin.graphs import Graph
+from netin.utils.constants import CLASS_ATTRIBUTE
+from netin.models import PAHModel
+
+from sympy import Eq
+from sympy import solve
+from sympy import symbols
+import powerlaw
+
+import numpy as np
+
+def cmp_new(a_h: np.ndarray, a_pa: np.ndarray) -> np.ndarray:
+    a_p = (a_h / a_h.sum()) * (a_pa / a_pa.sum())
+    return a_p / a_p.sum()
+
+def cmp_old(a_h: np.ndarray, a_pa: np.ndarray) -> np.ndarray:
+    a_p = a_h * a_pa
+    return a_p / a_p.sum()
+
+def calculate_edge_type_counts(graph: Graph) -> Dict[str, int]:
+    minority_nodes = graph.get_node_class(CLASS_ATTRIBUTE).get_class_values()
+    counter = {
+        "MM": 0,
+        "Mm": 0,
+        "mM": 0,
+        "mm": 0
+    }
+    for source, target in graph.edges():
+        counter[f"{minority_nodes[source]}{minority_nodes[target]}"] += 1
+    return counter
+
+def calculate_degree_powerlaw_exponents(graph: Graph) -> Tuple[float, float]:
+        degrees = graph.degrees()
+        minority_nodes = graph.get_node_class(CLASS_ATTRIBUTE)
+
+        dM = degrees[minority_nodes.get_majority_mask()]
+        dm = degrees[minority_nodes.get_minority_mask()]
+
+        fit_M = powerlaw.Fit(data=dM, discrete=True, xmin=min(dM), xmax=max(dM), verbose=False)
+        fit_m = powerlaw.Fit(data=dm, discrete=True, xmin=min(dm), xmax=max(dm), verbose=False)
+
+        pl_M = fit_M.power_law.alpha
+        pl_m = fit_m.power_law.alpha
+        return pl_M, pl_m
+
+def infer_homophily_values(graph: Graph) -> Tuple[float, float]:
+    """
+    Infers analytically the homophily values for the majority and minority classes.
+
+    Returns
+    -------
+    h_MM: float
+        homophily within majority group
+
+    h_mm: float
+        homophily within minority group
+    """
+
+    f_m = np.mean(graph.get_node_class(CLASS_ATTRIBUTE))
+    f_M = 1 - f_m
+
+    e = calculate_edge_type_counts(graph)
+    e_MM = e['MM']
+    e_mm = e['mm']
+    M = e['MM'] + e['mm'] + e['Mm'] + e['mM']
+
+    p_MM = e_MM / M
+    p_mm = e_mm / M
+
+    pl_M, pl_m = calculate_degree_powerlaw_exponents(graph)
+    b_M = -1 / (pl_M + 1)
+    b_m = -1 / (pl_m + 1)
+
+    # equations
+    hmm, hMM, hmM, hMm = symbols('hmm hMM hmM hMm')
+    eq1 = Eq((f_m * f_m * hmm * (1 - b_M)) / ((f_m * hmm * (1 - b_M)) + (f_M * hmM * (1 - b_m))), p_mm)
+    eq2 = Eq(hmm + hmM, 1)
+
+    eq3 = Eq((f_M * f_M * hMM * (1 - b_m)) / ((f_M * hMM * (1 - b_m)) + (f_m * hMm * (1 - b_M))), p_MM)
+    eq4 = Eq(hMM + hMm, 1)
+
+    solution = solve((eq1, eq2, eq3, eq4), (hmm, hmM, hMM, hMm))
+    h_MM, h_mm = solution[hMM], solution[hmm]
+    return h_MM, h_mm
+
+def main():
+    n = 2000
+    k = 2
+    f_m = .2
+    h = .8
+
+    a_h_M_infer, a_h_m_infer = [], []
+    for i in range(50):
+        model = PAHModel(n=n, f_m=f_m, k=k, h_mm=h, h_MM=h, seed=i)
+        g = model.simulate()
+
+        print(calculate_edge_type_counts(g))
+        h_M_inf, h_m_inf = infer_homophily_values(g)
+
+        a_h_M_infer.append(h_M_inf)
+        a_h_m_infer.append(h_m_inf)
+
+    a_h_M_infer = np.asarray(a_h_M_infer)
+    a_h_m_infer = np.asarray(a_h_m_infer)
+
+    print("Minority")
+    print(np.mean(a_h_m_infer))
+
+    print("Majority")
+    print(np.mean(a_h_M_infer))
+
+    a_h = np.asarray([.2, .2, .8, .2, .8])
+    a_pa = np.asarray([4, 8, 2, 2, 3])
+    print(
+        cmp_new(a_h, a_pa),
+        cmp_old(a_h, a_pa)
+    )
+
+if __name__ == "__main__":
+    main()