Flow control model training method and device, computer device and storage medium

Abstract

The application relates to a flow control model training method, which comprises the following steps: for each offline period, generating offline sample data of a next offline period based on offline sample data of a current offline period and offline flow control prediction data output by a basic flow control model; determining offline cumulative reward reference values of each offline period according to offline sample data of each offline period; performing offline period rolling training on the basic flow control model based on offline sample data and offline cumulative reward reference values of multiple offline periods to obtain an intermediate flow control model; for each online period, determining online cumulative reward reference values of a current online period through online sample data of the current online period; performing online period rolling training on the intermediate flow control model based on online sample data and online cumulative reward reference values of multiple online periods to obtain a target flow control model suitable for flow control data prediction in a multimedia communication process. The target flow control model adopting the method can improve flow control accuracy.

Images