Markov Network Structure Learning: A Randomized Feature Generation Approach
Haaren, Jan Van (KU Leuven - University of Leuven) | Davis, Jesse (KU Leuven - University of Leuven)
The structure of a Markov network is typically learned in one of two ways. The first approach is to treat this task as a global search problem. However, these algorithms are slow as they require running the expensive operation of weight (i.e., parameter) learning many times. The second approach involves learning a set of local models and then combining them into a global model. However, it can be computationally expensive to learn the local models for datasets that contain a large number of variables and/or examples. This paper pursues a third approach that views Markov network structure learning as a feature generation problem. The algorithm combines a data-driven, specific-to-general search strategy with randomization to quickly generate a large set of candidate features that all have support in the data. It uses weight learning, with L1 regularization, to select a subset of generated features to include in the model. On a large empirical study, we find that our algorithm is equivalently accurate to other state-of-the-art methods while exhibiting a much faster run time.
Jul-21-2012
- Country:
- Europe > Belgium
- Flanders > Flemish Brabant > Leuven (0.04)
- Asia > Middle East
- Jordan (0.04)
- Europe > Belgium
- Genre:
- Research Report (0.34)
- Technology: