| Welcome to |
|
|
business <> help for bidders <> business
| We have developed an
algorithm that can give bidders a better deal at auctions. The
algorithm
learns the preferences of the other bidders during an auction. It uses
neural networks, genetic algorithms, and fuzzy logic, three fields of
advanced computing that Prof. De Wilde has researched and taught for
ten
years at the top engineering school in the United Kingdom. We have been
able to combine the three fields successfully in an algorithm, and
during a sabbatical at UC Berkeley in 1999, Prof. De Wilde became aware
of its commercial potential. The algorithm will work for any type of
auction, as long as the bidders can modify their bids. There is
currently no competition for our algorithm. Most current auction advice
is based on valuation of goods rather than modelling behaviour of the
other bidders.
The market for this application is
business-to-business auctions. This is a rapidly growing market for
goods of significant value. The high end of the market for this
algorithm includes radio spectrum auctions and bond auctions, but we
are
happy to advise on any business-to-business auction. |
|
Information about the algorithm
The algorithm is based on game theory. More specifically, it uses the theory of soft games that we have developed in recent years. This works as follows. If you have to bid at an auction, you have numerical information (turnover, balance sheet) as well as non-numerical information (strong, weak, cautious) about your competitors that are also bidding in the auction. Our algorithm is the only one to use fuzzy logic to deal with this non-numerical information. Fuzzy logic expresses degrees of membership to a class, for example the class of strong competitors. It is very close to everyday language. It uniquely allows us to capture ALL information, numerical and non-numerical, that you have about your competitors.A second component of soft games is learning. In most b2b auctions you can revise your bids. Between two subsequent bids, we implement neural network learning. This is learning of patterns, just as the human brain learns patterns. The patterns we learn are the winning patterns. A pattern is winning for you if it allows you to get the good you want, for the lowest price. Our neural network learning is unique in that it can learn patterns that may be fuzzy in some components. It can deal with uncertainty about your competitors, for example when the auctioneer does not tell you the bids of your competitors.
The third component of soft games is genetic programing, or evolutionary games. During the auction, we run mock auctions, where the successful bidders multiply, and the unsuccessful ones disappear. In the end, we retain the strategy of the most successful bidder.
This survival of the fittest is combined with neural network learning. This is why we can claim to surpass human intelligence: your genetic makeup does not pass on to your offspring what you have learned during your lifetime, but our software can do exactly this during the mock auctions.
The combination of neural network and
genetic learning is fast. When we provide on-line advice, we can advise
you on the next bid in five minutes.