	US 12,135,741 B2
	Markov decision process for efficient data transfer
Reetesh Mukul, Bangalore (IN); and Mayuri Jain, Bangalore (IN)
Assigned to Adobe Inc., San Jose, CA (US)
Filed by Adobe Inc., San Jose, CA (US)
Filed on Jul. 30, 2020, as Appl. No. 16/943,331.
Prior Publication US 2022/0035855 A1, Feb. 3, 2022
Int. Cl. G06N 20/00 (2019.01); G06F 16/58 (2019.01)

CPC G06F 16/58 (2019.01) [G06N 20/00 (2019.01)]

20 Claims

1. A method for transferring a plurality of image files, the method comprising:

initiating a fetch operation on a batch of images having batch size of m_F, of the plurality of image files, wherein the fetch operation is one fetch operation of N fetch operations occurring in parallel with one another;

initiating an image processing operation on a batch of images having batch size of m_P, of the plurality of image files, wherein the image processing operation is one image processing operation of M image processing operations occurring in parallel with one another;

initiating a database update operation K_Don a batch of images having batch size of m_D, of the plurality of image files, wherein the fetch operation, the image processing operation, and the database update operation execute concurrently;

while the fetch operation, the image processing operation, and the database update operation are executing concurrently, using a Markov decision process to determine updated values of N, M, m_F, m_P, and m_Dat a time t_n, wherein executing the fetch operation, the image processing operation, and the database update operation using the updated values results in a higher reward compared to a previous time t_n−1, wherein the higher reward is associated with a lower transfer time for the plurality of image files compared to the previous time t_n−1and a lower average processing time for the image processing operation and the database update operation compared to the previous time t_n−1; and

concurrently executing the fetch operation, the image processing operation, and the database update operation using the updated values at a later time t_n+1.