我正在對神經網絡進行超參數調整。我嘗試了很多手動調整，但對我一直使用的數據集的預測能力仍然很差。我一直選擇使用 GridSearch 來測試我的模型的所有可能的參數組合。類似的事情是否可能（參見下面的代碼）或者是否有更智能/更好的參數調整方法？代碼能夠運行；當然這需要一些時間，但確實有效。我沒有特別的錯誤，我只是在尋找一些好的見解來知道這是否合適。數據框示例：sequence                                         target     expression-AQSVPWGISRVQAPAAH-NRGLRGSGVKVAVLDTGI-STHPDLNI...   0.00    50.0-AQQVPYGVSQIKAPALH-EQGYTGQNVKVAVIDTGIDSSHPDLKV...   0.46    42.0-AQSVPWGIRRVQAPAAH-NRGLTGSGVKVAVLDTGI-STHPDLNI...   0.34    46.0-AQTVPWGISRVQAPAAH-NRGLTGAGVKVSVLDTGI-STHPDLNI...   0.95    45.0-AQSVPYGVSQIKAPALH-SQGYTGSNVKVAVIDTGIDSSHPDLKV...   0.60    50.0數據形狀：3000行和3840個特征請注意，特征數很高，因為所有這些序列都是一次熱編碼的。代碼：'Hyperparameter Tuning for Neurons, Batch_Size, Epochs and Learning Rate'def build_regressor(n_neurons=1, learning_rate=0.01):    regressor = Sequential()    regressor.add(Dense(n_neurons, activation = 'relu', input_shape = (x_train.shape[1],)))    #regressor.add(Dense(n_neurons, activation = 'relu'))    regressor.add(Dense(units=1))    optimizer = Adam(lr = learning_rate)    regressor.compile(optimizer= optimizer, loss='mean_squared_error',  metrics=['mae','mse'])    return regressor#Create Modelmodel = KerasRegressor(build_fn=build_regressor, verbose=0)# define the grid search parametersbatch_size = [10, 25, 50, 100, 150]epochs = [5, 10, 25, 50]n_neurons = [1, 32, 64, 128, 256, 512]learning_rate = [0.001, 0.01, 0.1, 0.2, 0.3]param_grid = dict(batch_size=batch_size, epochs=epochs, n_neurons=n_neurons, learning_rate = learning_rate)#implement grid_searchgrid = GridSearchCV(estimator=model, param_grid=param_grid, n_jobs=-1, cv=3, scoring = 'r2')grid_result = grid.fit(x_train, y_train)# summarize resultsprint("Best: %f using %s" % (grid_result.best_score_, grid_result.best_params_))means = grid_result.cv_results_['mean_test_score']stds = grid_result.cv_results_['std_test_score']params = grid_result.cv_results_['params']for mean, stdev, param in zip(means, stds, params):    print("%f (%f) with: %r" % (mean, stdev, param))