CSE472-blanket-challenge/phase1-dataset

Phase 1 Dataset

Data Card

Field	Type	Description
data_id	string	Unique identifier (hash of graph_id + target index)
graph_id	string	Source graph identifier
num_nodes	int	Number of variables
num_edges	int	Number of edges in DAG
density	float	Edge density (edges / max_possible_edges)
target_variable	int	Index of target variable (non-root node)
adjacency_matrix	List[List[int]]	NxN binary matrix where A[i,j]=1 means i→j
markov_blanket	List[int]	Binary mask indicating Markov blanket members
data	List[List[float]]	Observational data (200 samples × num_nodes)
n_samples	int	Number of samples per dataset
method	string	Graph generation method (e.g., "ER", "SF")

Dataset Generation Settings

Graphs: Random DAGs with varying densities

Target Selection: Random non-root nodes

• Excludes root nodes (nodes with no incoming edges)
• Ensures target has causal dependencies

Data Generation: Linear Gaussian SCM

• For each node i in topological order: $X_i = \sum_j(\beta_{ij} * X_j) + \epsilon$
• Coefficients $\beta_{ij}$ sampled from [-1, 1]
• Noise from $N(0, 0.5^2)$
• 200 samples per dataset

Quick Start

Load Dataset

from datasets import load_dataset
import numpy as np

# Load dataset
dataset = load_dataset("json", data_files="phase1_dataset.jsonl")
record = dataset["train"][0]

print(f"Data shape: {len(record['data'])} x {len(record['data'][0])}")  # 200 x num_nodes
print(f"Target: {record['target_variable']}")
print(f"Graph density: {record['density']}")

Convert Adjacency Matrix to Graph

import networkx as nx
import numpy as np

# Get adjacency matrix
adj_matrix = np.array(record["adjacency_matrix"])

# Create directed graph
dag = nx.from_numpy_array(adj_matrix, create_using=nx.DiGraph)

print(f"Nodes: {dag.number_of_nodes()}")
print(f"Edges: {dag.number_of_edges()}")
print(f"Is DAG: {nx.is_directed_acyclic_graph(dag)}")

Extract Data and Markov Blanket

# Extract data and target
X = np.array(record["data"])  # (200, num_nodes)
target_idx = record["target_variable"]
y = X[:, target_idx]

# Get Markov blanket indices
mb_mask = np.array(record["markov_blanket"])
mb_indices = np.where(mb_mask == 1)[0]

print(f"Markov blanket members: {mb_indices}")
print(f"Data shape: {X.shape}")
print(f"Target shape: {y.shape}")