Implement fit/predict methods, unify subpackages, add tests

hyperbench/__init__.py

@@ -0,0 +1,11 @@
+from . import hyperlink_prediction
+from . import negative_sampling
+from . import pipelines
+from . import utils
+
+__all__ = [
+    "hyperlink_prediction", 
+    "negative_sampling", 
+    "pipelines", 
+    "utils"
+]

hyperbench/hyperlink_prediction/__init__.py

@@ -0,0 +1,7 @@
+from . import datasets, loader, models
+
+__all__ = [
+    "datasets",
+    "loader",
+    "models"
+]

hyperlink_prediction/datasets/__init__.py → hyperbench/hyperlink_prediction/datasets/__init__.py

@@ -1,10 +1,11 @@
 from .arb_dataset import ARBDataset
 from .dataset_hypergraph import DatasetHyperGraph
-from .imdb_dataset import IMDBHypergraphDataset, ARXIVHypergraphDataset, COURSERAHypergraphDataset
+from .imdb_dataset import IMDBHypergraphDataset, ARXIVHypergraphDataset, COURSERAHypergraphDataset, CHLPBaseDataset
 
 __all__ = data_classes = [
     'DatasetHyperGraph',
     'ARBDataset',
+    'CHLPBaseDataset',
     'IMDBHypergraphDataset',
     'COURSERAHypergraphDataset',
     'ARXIVHypergraphDataset'

hyperlink_prediction/datasets/arb_dataset.py → hyperbench/hyperlink_prediction/datasets/arb_dataset.py

@@ -2,7 +2,7 @@
 import tarfile
 import torch
 import torch.nn.functional as F
-from hyperlink_prediction.datasets.dataset_hypergraph import DatasetHyperGraph
+from .dataset_hypergraph import DatasetHyperGraph
 from torch_geometric.data.hypergraph_data import HyperGraphData
 from os import remove, listdir
 
@@ -22,7 +22,7 @@ class ARBDataset(DatasetHyperGraph):
         "tags-math-sx": "1eDevpF6EZs19rLouNpiKGLIlFOLUfKKG",
         "contact-high-school": "1VA2P62awVYgluOIh1W4NZQQgkQCBk-Eu",
         "contact-primary-school": "1sBHSEIyvVKavAho524Ro4cKL66W6rn-t",
-        "NDC-substances": "1dLJt3qzAOYieay03Sp9h8ZfVMiU-nMqC"
+        "NDC-substances": "1mGOg0DMh46J2zQdimSXMde1pKNtfAdh8"
     }
 
 

hyperlink_prediction/datasets/imdb_dataset.py → hyperbench/hyperlink_prediction/datasets/imdb_dataset.py

@@ -1,7 +1,7 @@
 import torch
 import pickle
 from abc import ABC
-from hyperlink_prediction.datasets.dataset_hypergraph import DatasetHyperGraph
+from .dataset_hypergraph import DatasetHyperGraph
 from torch_geometric.data.hypergraph_data import HyperGraphData
 
 

hyperlink_prediction/loader/dataloader.py → hyperbench/hyperlink_prediction/loader/dataloader.py

@@ -1,10 +1,10 @@
 import torch
 from typing import List, Any
 from torch.utils.data import DataLoader
-from utils.set_negative_samplig_method import setNegativeSamplingAlgorithm
-from negative_sampling.hypergraph_negative_sampling_algorithm import HypergraphNegativeSampler, MotifHypergraphNegativeSampler
+from ...utils.data_and_sampling_selector import setNegativeSamplingAlgorithm
+from ...negative_sampling.hypergraph_negative_sampling_algorithm import HypergraphNegativeSampler, MotifHypergraphNegativeSampler
 from torch_geometric.data.hypergraph_data import HyperGraphData
-from hyperlink_prediction.datasets.dataset_hypergraph import DatasetHyperGraph
+from ..datasets.dataset_hypergraph import DatasetHyperGraph
 
 
 class DatasetLoader(DataLoader):

hyperbench/hyperlink_prediction/models/__init__.py

@@ -0,0 +1,9 @@
+from .hyperlink_prediction_base import HyperlinkPredictor
+from .hyperlink_prediction_algorithm import CommonNeighbors
+from .hyperlink_prediction_result import HyperlinkPredictionResult
+
+__all__ = data_classes = [
+    'HyperlinkPredictor',
+    'CommonNeighbors',
+    'HyperlinkPredictionResult'
+]

hyperbench/hyperlink_prediction/models/hyperlink_prediction_algorithm.py

@@ -0,0 +1,45 @@
+import torch
+from torch import Tensor
+from .hyperlink_prediction_base import HyperlinkPredictor
+from .hyperlink_prediction_result import HyperlinkPredictionResult
+
+class CommonNeighbors(HyperlinkPredictor):
+    def __init__(self, device='cpu'):
+        super().__init__(device)
+        self.H = None        
+        self.num_node = None
+        self.num_hyperlink = None
+
+    def fit(self, X, y, edge_index, *args, **kwargs):
+
+        self.num_node = int(edge_index[0].max().item()) + 1
+        self.num_hyperlink = int(edge_index[1].max().item()) + 1
+
+        sparse = torch.sparse_coo_tensor(
+            edge_index,
+            torch.ones(edge_index.shape[1], device=self.device),
+            (self.num_node, edge_index.max().item() + 1),
+            device=self.device
+        )
+
+        self.H = sparse.to_dense()
+        return self 
+
+    def score_CN(self, H, u, v):
+        return torch.dot(H[u], H[v]).item()
+
+    def predict(self, edge_index: Tensor):
+        if self.H is None:
+            if edge_index is None:
+                raise ValueError("Model not fitted. Call fit() first.")
+            self.fit(None, None, edge_index)
+        H = self.H
+
+        CN_matrix = torch.matmul(H, H.T)
+
+        new_edges = torch.nonzero(torch.triu(CN_matrix, diagonal=1)).T 
+
+        return HyperlinkPredictionResult(
+            edge_index=new_edges,
+            device=self.device
+        )

hyperbench/hyperlink_prediction/models/hyperlink_prediction_base.py

@@ -0,0 +1,18 @@
+import torch
+import numpy as np
+from abc import abstractmethod
+from .hyperlink_prediction_result import HyperlinkPredictionResult
+class HyperlinkPredictor():
+
+    def __init__(self, num_node: int, device: torch.device = torch.device('cpu')):
+        self.num_node = num_node
+        self.device = device
+
+    @abstractmethod
+    def fit(self, X, y, edge_index, *args, **kwargs):
+        pass
+
+    @abstractmethod
+    def predict(self, X, edge_index: torch.Tensor) -> HyperlinkPredictionResult:
+        pass
+

negative_sampling/__init__.py → hyperbench/negative_sampling/__init__.py

@@ -1,4 +1,3 @@
-import negative_sampling
 from .hypergraph_negative_sampling import HypergraphNegativeSampler
 from .hypergraph_negative_sampling_result import HypergraphNegativeSamplerResult, ABSizedHypergraphNegativeSamplerResult
 from .hypergraph_negative_sampling_algorithm import ABSizedHypergraphNegativeSampler,SizedHypergraphNegativeSampler, MotifHypergraphNegativeSampler, CliqueHypergraphNegativeSampler

negative_sampling/hypergraph_negative_sampling_algorithm.py → hyperbench/negative_sampling/hypergraph_negative_sampling_algorithm.py

@@ -2,8 +2,8 @@
 import torch_geometric.nn.aggr as aggr
 from enum import Enum
 from torch import Tensor
-from negative_sampling.hypergraph_negative_sampling import HypergraphNegativeSampler
-from negative_sampling.hypergraph_negative_sampling_result import HypergraphNegativeSamplerResult, ABSizedHypergraphNegativeSamplerResult
+from .hypergraph_negative_sampling import HypergraphNegativeSampler
+from .hypergraph_negative_sampling_result import HypergraphNegativeSamplerResult, ABSizedHypergraphNegativeSamplerResult
 
 import warnings
 warnings.filterwarnings('ignore', category=UserWarning)
@@ -119,7 +119,7 @@ def generate(self, edge_index: Tensor) -> ABSizedHypergraphNegativeSamplerResult
             local_edge_index[1] += num_hyperedges
             num_hyperedges = torch.max(local_edge_index[1]) + 1
             negative_edge_index = torch.cat([negative_edge_index , local_edge_index], dim = 1)
-            global_positives = torch.cat([global_positives, probabilities], dim = 0)
+            global_positives = torch.empty((0, probabilities.shape[1]), dtype=torch.float32, device=self.device)
             global_replace_mask = torch.cat([global_replace_mask, replace_mask], dim = 0)
             global_replacement = torch.cat([global_replacement, replacement], dim = 0) 
 

negative_sampling/hypergraph_negative_sampling_result.py → hyperbench/negative_sampling/hypergraph_negative_sampling_result.py

@@ -2,7 +2,7 @@
 import torch_geometric.nn.aggr as aggr
 from abc import ABC
 from torch import Tensor
-from negative_sampling.hypergraph_negative_sampling import HypergraphNegativeSampler
+from .hypergraph_negative_sampling import HypergraphNegativeSampler
 
 class HypergraphNegativeSamplerResult(ABC):
     """ A class Result which rapresents the hypergraph generated by 

hyperbench/pipelines/__init__.py

@@ -0,0 +1,5 @@
+from . import pipeline
+
+__all__ = [
+    pipeline
+] 

pipelines/pipeline.py → hyperbench/pipelines/pipeline.py

@@ -2,7 +2,7 @@
 
 def execute():
     parser = argparse.ArgumentParser(description="Insert dataset_name, insert negative_sampling method")
-    parser.add_argument('--dataset_name', type=str, help="The dataset's name, possible dataset's name: \nIMDB,\nCURSERA,\nARXIV", required=True)
+    parser.add_argument('--dataset_name', type=str, help="The dataset's name, possible dataset's name: \nIMDB,\nCOURSERA,\nARXIV", required=True)
     parser.add_argument('--negative_sampling', type=str, help="negative sampling method to use, possible methods: \n SizedHypergraphNegativeSampler,\nMotifHypergraphNegativeSampler,\nCliqueHypergraphNegativeSampler", required=True)
     parser.add_argument('--hlp_method', type=str, help="hyperlink prediction method to use, possible method: \nCommonNeighbors", required=True)
     parser.add_argument('--output_path', type=str, help="Path to save the results", default="./results")
@@ -21,11 +21,10 @@ def execute():
     import matplotlib.pyplot as plt
     import time
     from random import randint, seed
-    from hyperlink_prediction.loader.dataloader import DatasetLoader
-    from hyperlink_prediction.hyperlink_prediction_algorithm import CommonNeighbors
-    from hyperlink_prediction.datasets.imdb_dataset import CHLPBaseDataset, IMDBHypergraphDataset, ARXIVHypergraphDataset, COURSERAHypergraphDataset
-    from utils.set_negative_samplig_method import setNegativeSamplingAlgorithm
-    from utils.hyperlink_train_test_split import train_test_split
+    from ..hyperlink_prediction.loader.dataloader import DatasetLoader
+    from ..hyperlink_prediction.models.hyperlink_prediction_algorithm import CommonNeighbors
+    from ..utils.data_and_sampling_selector import setNegativeSamplingAlgorithm, select_dataset
+    from ..utils.hyperlink_train_test_split import train_test_split
     from torch_geometric.nn import HypergraphConv
     from tqdm.auto import trange, tqdm
     from torch_geometric.data.hypergraph_data import HyperGraphData
@@ -57,14 +56,7 @@ def pre_transform(data: HyperGraphData):
 
         return data
 
-    dataset : CHLPBaseDataset
-    match(dataset_name):
-        case 'IMDB': 
-            dataset = IMDBHypergraphDataset("./data", pre_transform= pre_transform)
-        case 'ARXIV':
-            dataset = ARXIVHypergraphDataset("./data", pre_transform = pre_transform)
-        case 'COURSERA':
-            dataset = COURSERAHypergraphDataset("./data", pre_transform = pre_transform)
+    dataset = select_dataset(dataset_name, pre_transform= pre_transform)
 
     test_size = 0.2
     val_size = 0.0
@@ -170,7 +162,7 @@ def forward(self, x, x_e, edge_index):
             negative_test = negative_sampler.generate(h.edge_index)
 
             hlp_method = CommonNeighbors(h.num_nodes)
-            hlp_result = hlp_method.generate(negative_test.edge_index)
+            hlp_result = hlp_method.predict(negative_test.edge_index)
 
             y_pos = torch.ones(hlp_result.edge_index.size(1), 1)
             y_neg = torch.zeros(negative_test.edge_index.size(1), 1)

hyperbench/utils/__init__.py

@@ -0,0 +1,7 @@
+from .hyperlink_train_test_split import train_test_split
+from .data_and_sampling_selector import setNegativeSamplingAlgorithm, select_dataset
+__all__ = [
+    'train_test_split',
+    'setNegativeSamplingAlgorithm',
+    'select_dataset'
+]

hyperbench/utils/data_and_sampling_selector.py

@@ -0,0 +1,31 @@
+from ..hyperlink_prediction.datasets import ARBDataset, IMDBHypergraphDataset, ARXIVHypergraphDataset, COURSERAHypergraphDataset, CHLPBaseDataset
+from ..negative_sampling.hypergraph_negative_sampling_algorithm import SizedHypergraphNegativeSampler, MotifHypergraphNegativeSampler, CliqueHypergraphNegativeSampler, HypergraphNegativeSampler
+
+def setNegativeSamplingAlgorithm(ns_algorithm: str, num_node: int):
+    ns_method : HypergraphNegativeSampler
+    match(ns_algorithm):
+        case 'SizedHypergraphNegativeSampler':
+            ns_method = SizedHypergraphNegativeSampler(num_node)
+        case 'MotifHypergraphNegativeSampler': 
+            ns_method = MotifHypergraphNegativeSampler(num_node)
+        case 'CliqueHypergraphNegativeSampler':
+            ns_method = CliqueHypergraphNegativeSampler(num_node)
+
+    return ns_method
+
+def select_dataset(ds: str, pre_transform):
+
+    dataset : ARBDataset
+    if ds in ARBDataset.GDRIVE_IDs.keys():
+        dataset = ARBDataset(ds, pre_transform= pre_transform)
+    else:
+        dataset : CHLPBaseDataset
+        match(ds):
+            case 'IMDB': 
+                dataset = IMDBHypergraphDataset("./data", pre_transform= pre_transform)
+            case 'ARXIV':
+                dataset = ARXIVHypergraphDataset("./data", pre_transform = pre_transform)
+            case 'COURSERA':
+                dataset = COURSERAHypergraphDataset("./data", pre_transform = pre_transform)
+
+    return dataset

pyproject.toml

@@ -24,7 +24,7 @@ dependencies = [
 requires-python = ">=3.9"
 
 [project.scripts]
-pipeline = "pipelines.pipeline:execute"
+pipeline = "hyperbench.pipelines.pipeline:execute"
 
 [project.urls]
 Homepage = "https://www.isislab.it/"

setup.py

@@ -7,7 +7,7 @@
     include_package_data=True,
     entry_points={
         'console_scripts': [
-            'pipeline=pipelines.pipeline:execute',
+            'pipeline=hyperbench.pipelines.pipeline:execute',
         ],
     },
     install_requires=[],

tests/__init__.py

@@ -0,0 +1,5 @@
+from . import methods_test
+
+__all__ = [
+    "methods_test"
+]