Weird results using MLPClassifier in Python with sgd

Question

I've used the MLPClassifier with default values and the confusion matrix and its chart seems ok.

But when I try to change the "solver" option to "sgd" the results are terrible. In fact, they don't have any sense to me. Am I using something wrong withe MLPClassifier parameters?

My code:

from sklearn import datasets
from sklearn import model_selection
from sklearn.neural_network import MLPClassifier
from sklearn.metrics import accuracy_score
from sklearn.metrics import confusion_matrix
from sklearn import preprocessing
import pandas as pd


cancer = datasets.load_breast_cancer()
x = cancer.data
t = cancer.target
x_norm = preprocessing.scale(x)
tam_test = 0.30
semilla =7
xe, xt, te ,tt = model_selection.train_test_split(x, t, test_size = tam_test, random_state = semilla)

# IF I USE THIS STANDAR CLASSIFIER IS OK
clasificador = MLPClassifier(hidden_layer_sizes=(50,))  
# FROM HEREON NONE OF THEM SEEMS TO SHOW CORRECT RESULTS
# clasificador = MLPClassifier(hidden_layer_sizes=(50), max_iter=300,solver="sgd",learning_rate="adaptive", learning_rate_init=.005)
# clasificador =  MLPClassifier(hidden_layer_sizes=(200,), max_iter=1000, learning_rate_init=.05, solver ="sgd", learning_rate= "adaptive")
# clasificador = MLPClassifier(solver='sgd', hidden_layer_sizes=50,max_iter=150,random_state=0,activation='logistic',alpha=1e-5,learning_rate_init=0.2)

clasificador.fit(xe,te)
obtenido = clasificador.predict(xt)
clasificacion = accuracy_score(tt, obtenido)
print ("Clasificacion: %.4f" %(clasificacion))
matriz_de_confusion = confusion_matrix (tt, obtenido)
print (matriz_de_confusion)
pd.DataFrame(clasificador.loss_curve_).plot()`

You can see here a correct picture use defaults parameters

And here a couple of image use any of the non default parameters

I've tried to adjust different parameters with MLPClassifier to use SGD but the results seem wrong. I want to know how to use correctly SGD