# data.py — REFERENCE SOLUTION (PyTorch) import torch from torch.utils.data import DataLoader, TensorDataset import numpy as np def create_dataset(x, y, batch_size=32, shuffle=True): """Create a DataLoader from numpy arrays.""" x_t = torch.tensor(x, dtype=torch.float32) if isinstance(x, np.ndarray) else x.float() y_t = torch.tensor(y, dtype=torch.long) if isinstance(y, np.ndarray) else y.long() dataset = TensorDataset(x_t, y_t) return DataLoader(dataset, batch_size=batch_size, shuffle=shuffle) def normalize_features(x): """Normalize features to zero mean and unit variance.""" if isinstance(x, np.ndarray): x = torch.tensor(x, dtype=torch.float32) mean = x.mean(dim=0) std = x.std(dim=0) return (x - mean) / (std + 1e-7) def split_dataset(x, y, val_fraction=0.2): """Split data into train/val sets.""" if isinstance(x, np.ndarray): x = torch.tensor(x, dtype=torch.float32) if isinstance(y, np.ndarray): y = torch.tensor(y, dtype=torch.long) n = len(x) val_size = int(n * val_fraction) indices = torch.randperm(n) val_idx = indices[:val_size] train_idx = indices[val_size:] return (x[train_idx], y[train_idx]), (x[val_idx], y[val_idx])