I am new to RNNs and I want to build a one-to-many RNN,
usecase - given air speed for a particular city in the US, I want to know the air speed inside the city within the buildings.
#x_train
((a0,b0),
(a1,b1), (a2,b2), (a3,b3)....... (a100,b100))
## a = air speed(a lies between 0 and 100) in a city and b = air direction( b lies between 0 to 360) in a city
Using the following dependent variable
#Y_train
((x0,y0,z0),(x1,y1,z1),(x2,y2,z2),(x3,y3,z3),.......
(x10000,y10000,z10000))
#x0 = u component of air around building 1
#x1 = u component of air around building 2
#and so on
#y0 = v component of air around building 1
#y1 = v component of air around building 2 and so on
#similarly with z
There are 10,000 buildings in a city. given air speed and direction, air speed u,v,w components have to be found out around 10,000 buildings
My approach
I am trying to train a LSTM
model = Sequential()
model.add(LSTM(input_shape=(2,2),return_sequences=True, units=50))
model.add(Dropout(0.2))
model.add(LSTM(250,return_sequences=True))
model.add(Dropout(0.2))
model.add(TimeDistributed(Dense(3)))
model.add(Activation("linear"))
model.compile(loss="mse", optimizer="rmsprop")
model.fit(X, Y, epochs=10000, batch_size=1, verbose=2)
i am not sure of timesteps parameter and how to apply to this use case.
This question seems to be similar to this, It didn't help much though.
