Learning Auto-Bidding Agent with Transformer Model

Problem Define: In the case where all future traffics are known, the problem can be formulated as a constrained multi-objective optimization problems, By solving the above problem by Linear Programing, the optimal bidding formula can be derived as follows: bid=α0​⋅Vi​+α1​⋅C⋅Vi​ However, in practice, the auction environment is complex, with factors like uncertain online traffic, multiple ad slots with different expose rate, and fluctuating rank model coefficients.