def evaluate(self, train_index, test_index):
inner_pool = ThreadPool(self._algorithm_params["n_threads"])
async_result = {}
for i in range(self._algorithm_params["grid_search_folds"]):
async_result[i] = {}
outer_kernel = self._kernel[train_index, :][:, train_index]
y_train = self._y[train_index]
skf = StratifiedKFold(
n_splits=self._algorithm_params["grid_search_folds"], shuffle=True)
inner_cv = list(skf.split(np.zeros(len(y_train)), y_train))
for i in range(len(inner_cv)):
inner_train_index, inner_test_index = inner_cv[i]
inner_kernel = outer_kernel[
inner_train_index, :][:, inner_train_index]
x_test_inner = outer_kernel[inner_test_index, :][:,
inner_train_index]
y_train_inner, y_test_inner = (
y_train[inner_train_index],
y_train[inner_test_index],
)
for c in self._algorithm_params["c_range"]:
async_result[i][c] = inner_pool.apply_async(
self._grid_search,
(inner_kernel, x_test_inner, y_train_inner, y_test_inner,
c),
)
inner_pool.close()
inner_pool.join()
best_parameter = self._select_best_parameter(async_result)
x_test = self._kernel[test_index, :][:, train_index]
y_train, y_test = self._y[train_index], self._y[test_index]
_, y_hat, y_hat_train = self._launch_svc(outer_kernel, x_test, y_train,
y_test, best_parameter["c"])
result = dict()
result["best_parameter"] = best_parameter
result["evaluation"] = utils.evaluate_prediction_multiclass(
y_test, y_hat)
result["evaluation_train"] = utils.evaluate_prediction_multiclass(
y_train, y_hat_train)
result["y_hat"] = y_hat
result["y_hat_train"] = y_hat_train
result["y"] = y_test
result["y_train"] = y_train
result["y_index"] = test_index
result["x_index"] = train_index
return result
def _grid_search(self, kernel_train, x_test, y_train, y_test, c):
# y_hat is the value predicted
_, y_hat, _ = self._launch_svc(kernel_train, x_test, y_train, y_test, c)
res = utils.evaluate_prediction_multiclass(y_test, y_hat)
return res['balanced_accuracy']