Fast and Scalable Distributed Boolean Tensor Factorization

Park, Namyong; Oh, Seok-Jin; Kang, U

doi:10.1109/icde.2017.152

Cited by 17 publications

(19 citation statements)

References 25 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Several algorithms (e. g., [8,5,10]) approximately factorize n-way 0/1 tensors. The Boolean rank-r CAN-DECOMP/PARAFAC (CP) decomposition of a tensor T ∈ {0, 1} n i=1 Di aims to discover n matrices…”

Section: Heuristic Algorithmsmentioning

confidence: 99%

“…BCP ALS [8] heuristically seeks a good CP decomposition of a 0/1 tensor by Alternating Least Squares. DBTF [10] distributes that method on the Spark framework. Walk'n'Merge [5] discovers patterns through random walks in a graph: its vertices stand for the n-tuples with membership degrees at 1 and its edges link n-tuples that differ in one single dimension.…”

Section: Heuristic Algorithmsmentioning

confidence: 99%

“…They only need to be computed once, before running Algorithm 1. A greedy algorithm solves, in an exact way, Problem (10) under the (remaining) constraints (5), (6) and (8). Constraint (5) forces x e = 1 for all elements e involved in the fragment F .…”

Section: Algorithm 1: Bigfootmentioning

confidence: 99%

“…Walk'n'Merge [5] was provided by its authors. DBTF [10] is only distributed in binary form. All experiments are performed on a GNU/Linux system running on top of 2.4 GHz cores, 12 MB of cache and 32 GB of RAM.…”

Section: Algorithm 2: Forward-selectmentioning

confidence: 99%

“…TriclusterBox [9], Walk'n'Merge [5] and DBTF [10] only handle 0/1 tensors. To use them, every membership degree in the fuzzy tensor is rounded to 0 or 1.…”

Section: Real-world Tensormentioning

confidence: 99%

See 4 more Smart Citations

Bigfoot Climbing the Hill with ILP to Grow Patterns in Fuzzy Tensors

Maciel¹,

Alves²,

Cerf³

2019

Proceedings of the 2019 Conference of the International Fuzzy Systems Association and the European Society for Fuzzy Logic and

View full text Add to dashboard Cite

Fuzzy tensors encode to what extent n-ary predicates are satisfied. The disjunctive box cluster model is a regression model where sub-tensors are explanatory variables for the values in the fuzzy tensor. In this article, the most informative patterns according to that model, with high areas times squared densities, are grown by hill-climbing from fragments of them, that a complete algorithm provides. At every iteration, an optimization problem (or its linear relaxation) is solved thanks to integer linear programming (or greedily). A forward selection then chooses among the discovered patterns a non-redundant subset that fits, but does not overfit, the tensor. Experiments show the proposal discovers high-quality patterns and outperforms state-of-the-art approaches when applied to 0/1 tensors, a special case.

show abstract

Section: Heuristic Algorithmsmentioning

confidence: 99%

Section: Heuristic Algorithmsmentioning

confidence: 99%

Section: Algorithm 1: Bigfootmentioning

confidence: 99%

Section: Algorithm 2: Forward-selectmentioning

confidence: 99%

“…TriclusterBox [9], Walk'n'Merge [5] and DBTF [10] only handle 0/1 tensors. To use them, every membership degree in the fuzzy tensor is rounded to 0 or 1.…”

Section: Real-world Tensormentioning

confidence: 99%

See 3 more Smart Citations

Bigfoot Climbing the Hill with ILP to Grow Patterns in Fuzzy Tensors

Maciel¹,

Alves²,

Cerf³

2019

Proceedings of the 2019 Conference of the International Fuzzy Systems Association and the European Society for Fuzzy Logic and

View full text Add to dashboard Cite

show abstract

Fast and scalable method for distributed Boolean tensor factorization

Park

Kang

2019

The VLDB Journal

View full text Add to dashboard Cite

Time-aware tensor decomposition for sparse tensors

2021

View full text Add to dashboard Cite

Given a sparse time-evolving tensor, how can we effectively factorize it to accurately discover latent patterns? Tensor decomposition has been extensively utilized for analyzing various multi-dimensional real-world data. However, existing tensor decomposition models have disregarded the temporal property for tensor decomposition while most real-world data are closely related to time. Moreover, they do not address accuracy degradation due to the sparsity of time slices. The essential problems of how to exploit the temporal property for tensor decomposition and consider the sparsity of time slices remain unresolved. In this paper, we propose time-aware tensor decomposition (tatd), an accurate tensor decomposition method for sparse temporal tensors. tatd is designed to exploit time dependency and time-varying sparsity of real-world temporal tensors. We propose a new smoothing regularization with Gaussian kernel for modeling time dependency. Moreover, we improve the performance of tatd by considering time-varying sparsity. We design an alternating optimization scheme suitable for temporal tensor decomposition with our smoothing regularization. Extensive experiments show that tatd provides the state-of-the-art accuracy for decomposing temporal tensors.

show abstract

Fast and Scalable Distributed Boolean Tensor Factorization

Cited by 17 publications

References 25 publications

Bigfoot Climbing the Hill with ILP to Grow Patterns in Fuzzy Tensors

Bigfoot Climbing the Hill with ILP to Grow Patterns in Fuzzy Tensors

Fast and scalable method for distributed Boolean tensor factorization

Time-aware tensor decomposition for sparse tensors

Contact Info

Product

Resources

About