Our network will be outputting a single value, which is the scaled flow or expected change of the bitcoin price in some given minute based on the previous minutes.

We can get this output by using a single neuron. This neuron can be implemented in a Keras Dense Layer. It will take, as inputs, the output of multiple LSTM neurons, which we will cover in the next section. Lastly, the activation of this neuron can be tanh because we've scaled our data to the same scale as the hyperbolic tangent function, as seen here:

output = Dense(1, activation='tanh', name='output')(lstm2)