Learning in Short-Time Horizons with Measurable Costs

摘要: LEARNING IN SHORT-TIME HORIZONS WITH MEASURABLE COSTS Patrick B. Mullen Department of Computer Science Master Dynamic pricing is a difficult problem for machine learning. The environment noisy, dynamic and has measurable cost associated with exploration that necessitates learning be done in short-time horizons. These horizons force the algorithms to make decisions based on scarce data. In this work, various are compared context pricing. include Kalman filter, artificial neural networks, particle swarm optimization genetic algorithms. majority these have been modified handle problem. results show adaptations allow noisy conditions learn quickly.