This technology employs a belief propagation algorithm that improves the efficiency of matching advertisements and goods to consumers in online settings.
The ability to match potential consumers with advertisements and goods is a key component of marketing in the internet age. However, the diversity and scale of goods and consumers in the online marketplace creates an enormous optimization problem if an advertiser is to effectively reach their target audience. Current algorithms to address this problem utilize a linear, generalized bipartite matching approach that, while near 100% optimal, is to slow to be considered practical. As such, these programs only provide marginal optimization of advertisement matching.
This technology provides an algorithm based on belief propagation that solves generalized matching problems with high efficiency. Due to the distributable computations underlying belief propagation, this technology is capable of achieving near 100% optimal matching while remaining computationally efficient. As such, this technology promises to improve advertisers’ ability to match advertisements and goods to consumers in online settings.
This technology has been demonstrated to run hundreds of times faster than GOBLIN, a benchmark graph optimization package.
Patent Issued (US 9,117,235)
*IR M08-054
*Licensing Contact: Richard Nguyen