| import numpy as np |
| import random |
| import scipy.sparse as sp |
|
|
| class GraphAugmentor(object): |
| def __init__(self): |
| pass |
|
|
| @staticmethod |
| def node_dropout(sp_adj, drop_rate): |
| """Input: a sparse adjacency matrix and a dropout rate.""" |
| adj_shape = sp_adj.get_shape() |
| row_idx, col_idx = sp_adj.nonzero() |
| drop_user_idx = random.sample(range(adj_shape[0]), int(adj_shape[0] * drop_rate)) |
| drop_item_idx = random.sample(range(adj_shape[1]), int(adj_shape[1] * drop_rate)) |
| indicator_user = np.ones(adj_shape[0], dtype=np.float32) |
| indicator_item = np.ones(adj_shape[1], dtype=np.float32) |
| indicator_user[drop_user_idx] = 0. |
| indicator_item[drop_item_idx] = 0. |
| diag_indicator_user = sp.diags(indicator_user) |
| diag_indicator_item = sp.diags(indicator_item) |
| mat = sp.csr_matrix( |
| (np.ones_like(row_idx, dtype=np.float32), (row_idx, col_idx)), |
| shape=(adj_shape[0], adj_shape[1])) |
| mat_prime = diag_indicator_user.dot(mat).dot(diag_indicator_item) |
| return mat_prime |
|
|
| @staticmethod |
| def edge_dropout(sp_adj, drop_rate): |
| """Input: a sparse user-item adjacency matrix and a dropout rate.""" |
| adj_shape = sp_adj.get_shape() |
| edge_count = sp_adj.count_nonzero() |
| row_idx, col_idx = sp_adj.nonzero() |
| keep_idx = random.sample(range(edge_count), int(edge_count * (1 - drop_rate))) |
| user_np = np.array(row_idx)[keep_idx] |
| item_np = np.array(col_idx)[keep_idx] |
| edges = np.ones_like(user_np, dtype=np.float32) |
| dropped_adj = sp.csr_matrix((edges, (user_np, item_np)), shape=adj_shape) |
| return dropped_adj |
|
|
|
|
