You should attach right after a LSTM
layer a Dense
layer with as much neurons as you consider (that depends upon the LSTM
output itself), and on top of that one add a final Dense
classification layer with a single neuron, that'd be the binary output.
与恶龙缠斗过久,自身亦成为恶龙;凝视深渊过久,深渊将回以凝视…