Keras自动编码器，带网格搜索，用于降维

Question

我正在为高维数据的降维设计一个自动编码器。我已经使用Keras构建了一些简单的自动编码器模型来实现这个目标。但现在我想使用Sklearn的GridSearchCV来自动微调超参数

我用来评估模型的指标是后续t-sne的分类情况。我预计t-sne图上会出现两个聚类，并尝试用不同的分类值（例如疾病与正常、两性等）解释聚类

为此，我尝试了以下代码：

## To create the model
def create_model (optimizer = 'Adam', n_layer = 2, factorArray = [1000, 100], n_bottleneck = 10):
    n_inputs = X.shape[1]
    visible = Input(shape=(n_inputs,))
    e = visible
    i = 0
    j = 0
    for i in range(n_layer):
        e = Dense(factorArray[i])(e)
        e = BatchNormalization()(e)
        e = LeakyReLU()(e)
    bottleneck = Dense(n_bottleneck)(e)
    d = bottleneck
    factorArray = factorArray[::-1]    
    for j in range(n_layer):
        d = Dense(factorArray[j])(d)
        d = BatchNormalization()(d)
        d = LeakyReLU()(d)
    output = Dense(n_inputs, activation='linear')(d)
    model = Model(inputs = visible, outputs = output)
    model = model.compile(optimizer = optimizer, loss ='mse')
    return model

   model = KerasClassifier(build_fn=create_model)
   #model = KerasRegressor(build_fn=create_model)

    # compile autoencoder model 
    
    param_grid = {
              'optimizer': ['Adam'],
              'factorArray': [(1000,100)],
              'n_bottleneck' : [10],
              'epochs':[1,2],
              'batch_size':[128]
             }

    gs = GridSearchCV(
        estimator=model,
        param_grid=param_grid,
        scoring=make_scorer(silhouette_score),
        n_jobs=-1, 
        verbose=2,
        refit = 'false'
    )

    with warnings.catch_warnings(record=True) as w:
        try:
            gs.fit(X, y)  
        except ValueError:
            pass
        print(repr(w[-1].message))
    # fit the autoencoder model to reconstruct input
    history = gs.fit(X_train, X_train)
    # alternative 
    # history = gs.fit(X_train,X_train, validation = (X_test,X_test))
    df = pd.DataFrame(history.cv_results_)
    df.to_csv('Grid_search_table_' + key + ".csv")

程序引发了以下错误：

Traceback (most recent call last):
  File "/mnt/dzl_bioinf/exec/python_lib_keras/keraML_3.7/lib/python3.7/site-packages/sklearn/model_selection/_validation.py", line 593, in _fit_and_score
    estimator.fit(X_train, y_train, **fit_params)
  File "/mnt/dzl_bioinf/exec/python_lib_keras/keraML_3.7/lib/python3.7/site-packages/tensorflow/python/keras/wrappers/scikit_learn.py", line 223, in fit
    return super(KerasClassifier, self).fit(x, y, **kwargs)
  File "/mnt/dzl_bioinf/exec/python_lib_keras/keraML_3.7/lib/python3.7/site-packages/tensorflow/python/keras/wrappers/scikit_learn.py", line 159, in fit
    if (losses.is_categorical_crossentropy(self.model.loss) and
AttributeError: 'NoneType' object has no attribute 'loss'

  FitFailedWarning)
/mnt/dzl_bioinf/exec/python_lib_keras/keraML_3.7/lib/python3.7/site-packages/sklearn/model_selection/_validation.py:614: FitFailedWarning: Estimator fit failed. The score on this train-test partition for these parameters will be set to nan. Details: 
Traceback (most recent call last):
  File "/mnt/dzl_bioinf/exec/python_lib_keras/keraML_3.7/lib/python3.7/site-packages/sklearn/model_selection/_validation.py", line 593, in _fit_and_score
    estimator.fit(X_train, y_train, **fit_params)
  File "/mnt/dzl_bioinf/exec/python_lib_keras/keraML_3.7/lib/python3.7/site-packages/tensorflow/python/keras/wrappers/scikit_learn.py", line 223, in fit
    return super(KerasClassifier, self).fit(x, y, **kwargs)
  File "/mnt/dzl_bioinf/exec/python_lib_keras/keraML_3.7/lib/python3.7/site-packages/tensorflow/python/keras/wrappers/scikit_learn.py", line 159, in fit
    if (losses.is_categorical_crossentropy(self.model.loss) and
AttributeError: 'NoneType' object has no attribute 'loss'

Traceback (most recent call last):
  File "/var/spool/slurmd/job502071/slurm_script", line 267, in <module>
    history = gs.fit(X_train, X_train)
  File "/mnt/dzl_bioinf/exec/python_lib_keras/keraML_3.7/lib/python3.7/site-packages/sklearn/utils/validation.py", line 63, in inner_f
    return f(*args, **kwargs)
  File "/mnt/dzl_bioinf/exec/python_lib_keras/keraML_3.7/lib/python3.7/site-packages/sklearn/model_selection/_search.py", line 841, in fit
    self._run_search(evaluate_candidates)
  File "/mnt/dzl_bioinf/exec/python_lib_keras/keraML_3.7/lib/python3.7/site-packages/sklearn/model_selection/_search.py", line 1288, in _run_search
    evaluate_candidates(ParameterGrid(self.param_grid))
  File "/mnt/dzl_bioinf/exec/python_lib_keras/keraML_3.7/lib/python3.7/site-packages/sklearn/model_selection/_search.py", line 827, in evaluate_candidates
    _insert_error_scores(out, self.error_score)
  File "/mnt/dzl_bioinf/exec/python_lib_keras/keraML_3.7/lib/python3.7/site-packages/sklearn/model_selection/_validation.py", line 297, in _insert_error_scores
    raise NotFittedError("All estimators failed to fit")
sklearn.exceptions.NotFittedError: All estimators failed to fit

我的问题是：

1. 有没有人能帮我修复这个bug，或者告诉我如何在像我这样的无监督任务中使用GridSearchCV
2. 我是否有办法提取瓶颈层，并在运行网格搜索时将其可视化。附件中的手动培训和微调示例：

样本数据：

X1  X2  X3  X4  X5  

Label1  11.112964   8.633534    5.706326    8.432861    3.612596

Label5  9.02319 8.250181    5.636544    8.505976    3.488148

Label7  8.539526    7.338118    5.582104    8.626493    3.782542

Label8  9.622165    9.099065    5.855862    8.573742    3.466322