Discriminative Learning of Sum-Product Networks

By Robert Gens, Pedro M. Domingos. 2012

197 citations recorded by the source corpus.

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Citations in the corpus

Deep Learning: Methods and Applications (2014, background): “Another representative deep generative network that can be used for unsupervised (as well as supervised) learning is the sum–product network or SPN [125, 289].”
Convolutional Kernel Networks (2014, neutral): “Method [12] [27] [18] [13] [4] [17] [31] CKN-GM CKN-PM CKN-CO CIFAR-10 82.”
Bilinear CNN Models for Fine-Grained Visual Recognition (2015, methodology): “support vector machines [55]. However, one might be able to achieve similar approximations using alternate architectures such as sum-product networks that efﬁciently model multiplicative interactions [22]. Bilinearmodelvariantsandextensions.Bilinear models were used by Tanenbaum and Freeman [62] to ”
A tutorial survey of architectures, algorithms, and applications for deep learning (2014, background): “Properties of “completeness” and “consistency” constrain the SPN in a desirable way.”
Extrapolating Learning Curves of Deep Neural Networks (2014, methodology): “We preprocessed images by extracting k-means features extracted as described by Coates et al. (2011) and also done by Gens and Domingos (2012) and Swersky et al. (2013)).”
Bilinear CNNs for Fine-grained Visual Recognition (2015, background): “However, one might be able to achieve similar approximations using alternate architectures such as sum-product networks that efficiently model multiplicative interactions [22].”
Improper Deep Kernels (2016, methodology): “We use the CIFAR-10 and STL-10 datasets, with the same preprocessing as in [7].”
Learning the Structure of Sum-Product Networks (2013, background): “For the discriminative case, Gens and Domingos (2012) proposed a backpropagation-style gradient descent algorithm which obtained state-of-the-art results on image classification problems, but again faced a trade-off between flexibility and cost of learning.”
Learning hierarchical sparse features for RGB-(D) object recognition (2014, methodology): “In addition, HMP outperforms discriminative sum-product networks (SPNs), a recently introduced deep learning model (Gens and Domingos, 2012) that is trained in a supervised manner using an efﬁcient backpropagation-style algorithm.”
Contractive Rectifier Networks for Nonlinear Maximum Margin Classification (2015, unknown): “Sum-product Networks [9] 16.”
Spiking Deep Networks with LIF Neurons (2015, background): “More recently, many groups have achieved better results on these small datasets (e.g. [5]) and as well as success on larger datasets (e.g. [6]).”
Deep Representation Learning with Target Coding (2015, unknown): “50 – (Gens and Domingos 2012): SPN – – – – 62.”
Online Algorithms for Sum-Product Networks with Continuous Variables (2016, background): “Keywords: Sum-product networks; continuous variables; online learning.”
A Deep Architecture for Matching Short Texts (2013, background): “Our work is in a sense related to the sum-product network (SPN)[4, 5, 15], especially the work in [4] that learns the deep architecture from clustering in the feature space for the image completion task.”
Multi-Task Bayesian Optimization (2013, methodology): “Overfitting is a significant challenge for this dataset, so we utilize a CNN2 on top of k-means features in a similar approach to [25], as well as dropout [26].”
Language modeling with sum-product networks (2014, background): “SPNs have been successfully applied to vision problems [1, 2], but to date, no one has brought it to bear on the problem of language modeling.”
A Deep Sum-Product Architecture for Robust Facial Attributes Analysis (2013, methodology): “Our SPN is more efficient and robust than the existing SPNs [17, 7], which use pixel values or low-level features as input, because we operate on classifiers’ scores.”
Convex Deep Learning via Normalized Kernels (2014, background): “Other recent work [17] considers a sum-product formulation [18], where guarantees can be made about the efficient recovery of an approximately optimal polynomial basis.”