Deep Reinforcement Learning with a Combinatorial Action Space for Predicting Popular Reddit Threads
This work addresses the challenge of online popularity prediction for social media platforms, but it is incremental as it applies existing deep reinforcement learning techniques to a new benchmark task.
The paper tackles the problem of predicting popular Reddit threads using deep reinforcement learning with a combinatorial action space, achieving the best performance across different configurations and domains with a model that generalizes well to varying recommendation requests.
We introduce an online popularity prediction and tracking task as a benchmark task for reinforcement learning with a combinatorial, natural language action space. A specified number of discussion threads predicted to be popular are recommended, chosen from a fixed window of recent comments to track. Novel deep reinforcement learning architectures are studied for effective modeling of the value function associated with actions comprised of interdependent sub-actions. The proposed model, which represents dependence between sub-actions through a bi-directional LSTM, gives the best performance across different experimental configurations and domains, and it also generalizes well with varying numbers of recommendation requests.