rtb_da_meta_rl

Real Time Bidding in Display Advertising with Meta Reinforcement Learning

Start experiments

To start experiment to measure performance future actions (T-shoot learning):

$python bid_performance_on_future_auctions.py

To start experiment to measure performance on future campaigns:

$python bid_performance_new_campaigns.py

Adjust which methods to experiment on

Edit either bid_performance_on_future_auctions.py or bid_performance_new_campaigns.py and search for array agents_to_execute, there will be a set of potential agents.

The potential agent values are:

• meta_imitation: Meta RL with imitation learning as initialization step.
• lin: Linear agent
• rlb_rl_dp_tabular: Reinforcement Learning for bid based on dynamic programming.
• rlb_rl_fa: Reinforcement Learning for bid, based on function approximation with imitation learning as initialization step.