Removing autolguon as optional dependency

The autogluon models can still be used with DoWhy, but it is not listed as an optional dependency anymore to reduce the number of restrictions on other packages.

Signed-off-by: Patrick Bloebaum <bloebp@amazon.com>

Author: bloebp

docs/source/example_notebooks/dowhy_causal_discovery_example.ipynb

@@ -127,7 +127,7 @@
     "for method, lib in functions.items():\n",
     "    obj = lib()\n",
     "    output = obj.predict(data_mpg)\n",
-    "    adj_matrix = nx.to_numpy_matrix(output)\n",
+    "    adj_matrix = nx.to_numpy_array(output)\n",
     "    adj_matrix = np.asarray(adj_matrix)\n",
     "    graph_dot = make_graph(adj_matrix, labels)\n",
     "    graphs[method] = graph_dot\n",
@@ -252,7 +252,7 @@
    "outputs": [],
    "source": [
     "labels = [f'{col}' for i, col in enumerate(data_sachs.columns)]\n",
-    "adj_matrix = nx.to_numpy_matrix(graph_sachs)\n",
+    "adj_matrix = nx.to_numpy_array(graph_sachs)\n",
     "adj_matrix = np.asarray(adj_matrix)\n",
     "graph_dot = make_graph(adj_matrix, labels)\n",
     "display(graph_dot)"
@@ -293,7 +293,7 @@
     "    obj = lib()\n",
     "    output = obj.predict(data_sachs)\n",
     "    graphs_nx[method] = output\n",
-    "    adj_matrix = nx.to_numpy_matrix(output)\n",
+    "    adj_matrix = nx.to_numpy_array(output)\n",
     "    adj_matrix = np.asarray(adj_matrix)\n",
     "    graph_dot = make_graph(adj_matrix, labels)\n",
     "    graphs[method] = graph_dot\n",

docs/source/example_notebooks/dowhy_optimize_backdoor_example.ipynb

@@ -50,7 +50,7 @@
     "nodes = []\n",
     "for i in graph.nodes:\n",
     "    nodes.append(str(i))\n",
-    "adjacency_matrix = np.asarray(nx.to_numpy_matrix(graph))\n",
+    "adjacency_matrix = np.asarray(nx.to_numpy_array(graph))\n",
     "graph_dot = graph_operations.adjacency_matrix_to_graph(adjacency_matrix, nodes)\n",
     "graph_dot = graph_operations.str_to_dot(graph_dot.source)\n",
     "print(\"Graph Generated.\")\n",

docs/source/example_notebooks/gcm_supply_chain_dist_change.ipynb

@@ -89,7 +89,7 @@
    },
    "outputs": [],
    "source": [
-    "data.groupby(['week']).mean()[['received']].plot(kind='bar', title='average received', legend=False); "
+    "data.groupby(['week']).mean(numeric_only=True)[['received']].plot(kind='bar', title='average received', legend=False);"
    ]
   },
   {
@@ -130,7 +130,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "data.groupby(['week']).mean().plot(kind='bar', title='average', legend=True);"
+    "data.groupby(['week']).mean(numeric_only=True).plot(kind='bar', title='average', legend=True);"
    ]
   },
   {
@@ -305,7 +305,7 @@
     "data_week2 = buying_data(2, 1, demand_mean=4)\n",
     "data_week2['week'] = 'w2'\n",
     "\n",
-    "data = data_week1.append(data_week2, ignore_index=True)\n",
+    "data = pd.concat([data_week1, data_week2], ignore_index=True)\n",
     "# write data to a csv file\n",
     "# data.to_csv('supply_chain_week_over_week.csv', index=False)"
    ]

dowhy/causal_graph.py

@@ -443,7 +443,7 @@ def get_adjacency_matrix(self, *args, **kwargs):
         Get adjacency matrix from the networkx graph
 
         """
-        return nx.convert_matrix.to_numpy_matrix(self._graph, *args, **kwargs)
+        return nx.convert_matrix.to_numpy_array(self._graph, *args, **kwargs)
 
     def check_valid_frontdoor_set(
         self, nodes1, nodes2, candidate_nodes, frontdoor_paths=None, new_graph=None, dseparation_algo="default"

dowhy/causal_identifier/backdoor.py

@@ -101,7 +101,7 @@ def get_backdoor_vars(self):
 
         # Get adjacency list
         adjlist = adjacency_matrix_to_adjacency_list(
-            nx.to_numpy_matrix(undirected_graph), labels=list(undirected_graph.nodes)
+            nx.to_numpy_array(undirected_graph), labels=list(undirected_graph.nodes)
         )
         path_dict = {}
         backdoor_sets = []  # Put in backdoor sets format

dowhy/gcm/_noise.py

@@ -143,11 +143,11 @@ def noise_samples_of_ancestors(
             continue
 
         if is_root_node(causal_model.graph, node):
-            noise = causal_model.causal_mechanism(node).draw_samples(num_samples)
+            noise = causal_model.causal_mechanism(node).draw_samples(num_samples).reshape(-1)
             drawn_noise_samples[node] = noise
             drawn_samples[node] = noise
         else:
-            noise = causal_model.causal_mechanism(node).draw_noise_samples(num_samples)
+            noise = causal_model.causal_mechanism(node).draw_noise_samples(num_samples).reshape(-1)
             drawn_noise_samples[node] = noise
             drawn_samples[node] = causal_model.causal_mechanism(node).evaluate(
                 _parent_samples_of(node, causal_model, drawn_samples), noise

dowhy/gcm/fitting_sampling.py

@@ -92,9 +92,11 @@ def draw_samples(causal_model: ProbabilisticCausalModel, num_samples: int) -> pd
         causal_mechanism = causal_model.causal_mechanism(node)
 
         if is_root_node(causal_model.graph, node):
-            drawn_samples[node] = causal_mechanism.draw_samples(num_samples)
+            drawn_samples[node] = causal_mechanism.draw_samples(num_samples).reshape(-1)
         else:
-            drawn_samples[node] = causal_mechanism.draw_samples(_parent_samples_of(node, causal_model, drawn_samples))
+            drawn_samples[node] = causal_mechanism.draw_samples(
+                _parent_samples_of(node, causal_model, drawn_samples)
+            ).reshape(-1)
 
     return drawn_samples
 

dowhy/gcm/uncertainty.py

@@ -4,7 +4,7 @@
 """
 
 import numpy as np
-from numpy.dual import det
+from numpy.linalg import det
 from scipy.special import digamma
 from scipy.stats import entropy
 from sklearn.neighbors import NearestNeighbors

dowhy/gcm/validation.py

@@ -4,7 +4,7 @@
 the future.
 """
 from enum import Enum, auto
-from typing import Any, Callable, Dict, Optional, Set, Tuple, Union
+from typing import Any, Callable, Dict, List, Optional, Tuple, Union
 
 import networkx as nx
 import numpy as np
@@ -59,7 +59,7 @@ def refute_causal_structure(
         if parents and non_descendants:
             # test local Markov condition, null hypothesis: conditional independence
             lmc_p_value = conditional_independence_test(
-                data[node].values, data[non_descendants].values, data[parents].values
+                data[node].to_numpy(), data[non_descendants].to_numpy(), data[parents].to_numpy()
             )
             lmc_test_result = dict(p_value=lmc_p_value)
             all_p_values.append(lmc_p_value)
@@ -160,8 +160,8 @@ def refute_invertible_model(
         )
 
 
-def _get_non_descendants(causal_graph: DirectedGraph, node: Any, exclude_parents: bool = False) -> Set[Any]:
+def _get_non_descendants(causal_graph: DirectedGraph, node: Any, exclude_parents: bool = False) -> List[Any]:
     nodes_to_exclude = nx.descendants(causal_graph, node).union({node})
     if exclude_parents:
         nodes_to_exclude = nodes_to_exclude.union(causal_graph.predecessors(node))
-    return set(causal_graph.nodes).difference(nodes_to_exclude)
+    return list(set(causal_graph.nodes).difference(nodes_to_exclude))

dowhy/graph_learners/cdt.py

@@ -27,7 +27,7 @@ def learn_graph(self, labels=None):
         graph = self._method.predict(self._data)
 
         # Get adjacency matrix
-        self._adjacency_matrix = nx.to_numpy_matrix(graph)
+        self._adjacency_matrix = nx.to_numpy_array(graph)
         self._adjacency_matrix = np.asarray(self._adjacency_matrix)
 
         # If labels not provided