The following snippet shows the use of AdaGrad with Keras:
from keras.optimizers import Adagrad
...
adagrad = Adagrad(lr=0.0001, epsilon=1e-6, decay=1e-2)
model.compile(optimizer=adagrad,
loss='categorical_crossentropy',
metrics=['accuracy'])
The AdaGrad implementation has no other parameters but the common ones.