Inputs: 2 types of inputs are used. 1) Public Input which is the Market States for the stocks. At a time, 10 previous minutes of Market Information is nfed. 2) Private Input which is the Left Time and Left Executed Order
Model: The 2 types of inputs are fed into 2 RNN Networks. Their outputs are concatenated and fed into an Actor Critic Netwrok. The Output of the Actor is the fraction of volume to be traded at the next minute.