Abstract | ||
---|---|---|
Sum-Product Networks (SPNs) are a probabilistic graphical model with deep learning applications. A key feature in an SPN is that inference is linear with respect to the size of the network under certain structural constraints. Initial studies of SPNs have investigated transforming SPNs into Bayesian Networks (BNs). Two such methods modify the SPN before conversion. One method modifies the SPN into a normal form. The resulting BN does not contain edges between latent variables. The other method considered here augments the SPN with twin nodes. Here, the constructed BN does contain edges between latent variables, thereby encoding a richer set of dependencies among them. In this paper, we propose another method for converting an SPN into a BN. Our process starts with the normal SPN from the first method above. We introduce an augmented version of the normal SPN. Consequently, the constructed BN does contain edges between latent variables. The salient feature of our method is that, given a normal SPN, we build a BN not limited to a bipartite structure. Moreover, unlike a normal SPN, our augmented, normal SPN is necessarily complete. Lastly, by using aspects of two earlier methods, our approach can be seen as unifying the two methods. |
Year | DOI | Venue |
---|---|---|
2017 | 10.1007/978-3-319-57351-9_37 | ADVANCES IN ARTIFICIAL INTELLIGENCE, CANADIAN AI 2017 |
Keywords | Field | DocType |
Sum-product networks,Conditional independence,Deep learning | Pattern recognition,Computer science,Conditional independence,Inference,Theoretical computer science,Latent variable,Bayesian network,Artificial intelligence,Deep learning,Probabilistic logic,Graphical model,Encoding (memory) | Conference |
Volume | ISSN | Citations |
10233 | 0302-9743 | 0 |
PageRank | References | Authors |
0.34 | 3 | 3 |
Name | Order | Citations | PageRank |
---|---|---|---|
André E. dos Santos | 1 | 5 | 7.02 |
Cory J. Butz | 2 | 383 | 40.80 |
Jhonatan de S. Oliveira | 3 | 6 | 7.43 |