| dc.description.abstract | In the one-way trading problem, a player is presented with a number of rates. For each rate the player needs to decide if they want to trade, and how much of their budget they want to trade for a profit. This trade will be done according to the rate currently available to the player. The goal of the player is to maximise the profit they make on the sum of all trades. The discussed algorithm for one-way trading uses a semi-online environment. More specifically, the number of rates that will be presented is known and the bounds on the minimum and the maximum exchange rates are also known.
In this work, the online one-way trading algorithm is extended so that it can use a piece of advice that is given before the algorithm starts. This advice will influence the performance of the algorithm. More precisely, the effect on the performance will depend on the error of the advice and on the trust that is put in the advice.
With advice that has no error, the algorithm can achieve the optimal amount of profit. Depending on how big the error is, the algorithm that uses advice can perform worse than the original online algorithm that doesn't use advice. In this work, two different kinds of advice are tested. The best day to trade on and the highest rate that will be presented.
With the advice tested in this work, one-way trading with perfect advice will always be as good or better than one-way trading without advice. When the presented number of rates is large the online algorithm with advice can outperform the standard online algorithm, even with a significant error in the advice. This is especially true for the advice that predicts the best day to trade on. | |
