Deep Generalized Convolutional Sum-Product Networks

被引:0
|
作者
van de Wolfshaar, Jos [1 ,2 ]
Pronobis, Andrzej [1 ,3 ]
机构
[1] KTH Royal Inst Technol, Stockholm, Sweden
[2] MessageBird, Amsterdam, Netherlands
[3] Univ Washington, Seattle, WA 98195 USA
基金
瑞典研究理事会;
关键词
Sum-Product Networks; Deep Probabilistic Models; Image Representations;
D O I
暂无
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Sum-Product Networks (SPNs) are hierarchical, graphical models that combine benefits of deep learning and probabilistic modeling. SPNs offer unique advantages to applications demanding exact probabilistic inference over high-dimensional, noisy inputs. Yet, compared to convolutional neural nets, they struggle with capturing complex spatial relationships in image data. To alleviate this issue, we introduce Deep Generalized Convolutional Sum-Product Networks (DGC-SPNs), which encode spatial features in a way similar to CNNs, while preserving the validity of the probabilistic SPN model. As opposed to existing SPN-based image representations, DGC-SPNs allow for overlapping convolution patches through a novel parameterization of dilations and strides, resulting in significantly improved feature coverage and feature resolution. DGC-SPNs substantially outperform other SPN architectures across several visual datasets and for both generative and discriminative tasks, including image inpainting and classification. These contributions are reinforced by the first simple, scalable, and GPU-optimized implementation of SPNs, integrated with the widely used Keras/TensorFlow framework. The resulting model is fully probabilistic and versatile, yet efficient and straightforward to apply in practical applications in place of traditional deep nets.
引用
收藏
页码:533 / 544
页数:12
相关论文
共 50 条
  • [1] Deep Convolutional Sum-Product Networks
    Butz, Cory J.
    Oliveira, Jhonatan S.
    dos Santos, Andre E.
    Teixeira, Andre L.
    [J]. THIRTY-THIRD AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE / THIRTY-FIRST INNOVATIVE APPLICATIONS OF ARTIFICIAL INTELLIGENCE CONFERENCE / NINTH AAAI SYMPOSIUM ON EDUCATIONAL ADVANCES IN ARTIFICIAL INTELLIGENCE, 2019, : 3248 - 3255
  • [2] Sum-Product Networks: A New Deep Architecture
    Poon, Hoifung
    Domingos, Pedro
    [J]. 2011 IEEE INTERNATIONAL CONFERENCE ON COMPUTER VISION WORKSHOPS (ICCV WORKSHOPS), 2011,
  • [3] Sum-Product Networks: Deep Models with Tractable Inference
    Domingos, Pedro
    [J]. MACHINE LEARNING AND KNOWLEDGE DISCOVERY IN DATABASES, ECML PKDD 2015, PT I, 2015, 9284
  • [4] Sum-Product decoding of convolutional codes
    Shohon, Toshiyuki
    Ogawa, Yuuichi
    Ogiwara, Haruo
    [J]. IWSDA'09: PROCEEDINGS OF THE FOURTH INTERNATIONAL WORKSHOP ON SIGNAL DESIGN AND ITS APPLICATIONS IN COMMUNICATIONS, 2009, : 64 - +
  • [5] Residual Sum-Product Networks
    Ventola, Fabrizio
    Stelzner, Karl
    Molina, Alejandro
    Kersting, Kristian
    [J]. INTERNATIONAL CONFERENCE ON PROBABILISTIC GRAPHICAL MODELS, VOL 138, 2020, 138 : 545 - 556
  • [6] Sum-Product Networks: A Survey
    Sanchez-Cauce, Raquel
    Paris, Iago
    Javier Diez, Francisco
    [J]. IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, 2022, 44 (07) : 3821 - 3839
  • [7] Survey of sum-product networks
    Dai, Qi
    Liu, Jian-Wei
    [J]. Kongzhi Lilun Yu Yingyong/Control Theory and Applications, 2024, 41 (11): : 1965 - 1990
  • [8] Robustifying sum-product networks
    Maua, Denis Deratani
    Conaty, Diarmaid
    Cozman, Fabio Gagliardi
    Poppenhaeger, Katja
    de Campos, Cassio Polpo
    [J]. INTERNATIONAL JOURNAL OF APPROXIMATE REASONING, 2018, 101 : 163 - 180
  • [9] Sum-Product Autoencoding: Encoding and Decoding Representations Using Sum-Product Networks
    Vergari, Antonio
    Peharz, Robert
    Di Mauro, Nicola
    Molina, Alejandro
    Kersting, Kristian
    Esposito, Floriana
    [J]. THIRTY-SECOND AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE / THIRTIETH INNOVATIVE APPLICATIONS OF ARTIFICIAL INTELLIGENCE CONFERENCE / EIGHTH AAAI SYMPOSIUM ON EDUCATIONAL ADVANCES IN ARTIFICIAL INTELLIGENCE, 2018, : 4163 - 4170
  • [10] Mixed Sum-Product Networks: A Deep Architecture for Hybrid Domains
    Molina, Alejandro
    Vergari, Antonio
    Di Mauro, Nicola
    Natarajan, Sriraam
    Esposito, Floriana
    Kersting, Kristian
    [J]. THIRTY-SECOND AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE / THIRTIETH INNOVATIVE APPLICATIONS OF ARTIFICIAL INTELLIGENCE CONFERENCE / EIGHTH AAAI SYMPOSIUM ON EDUCATIONAL ADVANCES IN ARTIFICIAL INTELLIGENCE, 2018, : 3828 - 3835