Numpy数组形状与modAL和skLearn不匹配

Question

我遇到了numpy数组形状不匹配错误。在StackOverflow上有几十个这样的问题，但没有一个能帮助我解决这个问题。我正在努力调试，因为错误是由库代码引发的。下面是相关的部分：

from modAL.models import ActiveLearner

# ... fetch data

print("Beginning debugging logs:")
print(f"classifier: {clf}")
print(f"X_train shape: {X_train.shape}")
print(f"y_train shape: {y_train.shape}")
print(f"X_pool shape: {X_POOL.shape}")

learner = ActiveLearner(
    estimator=clf,
    X_training=X_train,
    y_training=y_train
    )

result = learner.query(X_POOL)

下面是回溯：

classifier: DecisionTreeClassifier(max_depth=4)
X_train shape: (84, 4926)
y_train shape: (84, 51)
X_pool shape: (997, 4926)


  File "rpc_server.py", line 139, in <module>
    result = learner.query(X_POOL)
  File "/home/ubuntu/venv/lib/python3.8/site-packages/modAL/models/base.py", line 261, in query
    query_result = self.query_strategy(self, X_pool, *query_args, **query_kwargs)
  File "/home/ubuntu/venv/lib/python3.8/site-packages/modAL/uncertainty.py", line 152, in uncertainty_sampling
    uncertainty = classifier_uncertainty(classifier, X, **uncertainty_measure_kwargs)
  File "/home/ubuntu/venv/lib/python3.8/site-packages/modAL/uncertainty.py", line 82, in classifier_uncertainty
    uncertainty = 1 - np.max(classwise_uncertainty, axis=1)
  File "/home/ubuntu/venv/lib/python3.8/site-packages/numpy/core/fromnumeric.py", line 2504, in amax
    return _wrapreduction(a, np.maximum, 'max', axis, None, out, keepdims=keepdims,
  File "/home/ubuntu/venv/lib/python3.8/site-packages/numpy/core/fromnumeric.py", line 86, in _wrapreduction
    return ufunc.reduce(obj, axis, dtype, out, **passkwargs)
ValueError: could not broadcast input array from shape (997,1) into shape (997)

该环境为Python3.8，具有：

pandas==1.1.4
numpy==1.16.0
sklearn==0.0
modAL==0.4.0

我相信(997,1)只是一个普通的嵌套(997)数组？它似乎离它所需要的并不太远。

如何调试此错误？谢谢!

Answer 1

您正在接收的ValueError似乎与线路有关：

uncertainty = 1 - np.max(classwise_uncertainty, axis=1)

也许您可以使用以下命令完全避免在ufunc.reduce级别进行广播：

uncertainty = 1 - (np.max(classwise_uncertainty, axis=1, keepdims=1)).flatten()

但是我应该注意，我无法在我的机器上重现这个错误。在最坏的情况下，您可以简单地手动重新制定代码行(使用一两个循环)，以反映它所编码的相同数学。