Filters








19,586 Hits in 5.9 sec

Approximate Inference in Structured Instances with Noisy Categorical Observations [article]

Alireza Heidari, Ihab F. Ilyas, Theodoros Rekatsinas
2019 arXiv   pre-print
We study the problem of recovering the latent ground truth labeling of a structured instance with categorical random variables in the presence of noisy observations.  ...  We present a new approximate algorithm for graphs with categorical variables that achieves low Hamming error in the presence of noisy vertex and edge observations.  ...  Approximate Inference in Structured Instances with Noisy Categorical Observations  ... 
arXiv:1907.00141v2 fatcat:dainjewajnhsjjkw6e6so22mse

Deperturbation of Online Social Networks via Bayesian Label Transition [article]

Jun Zhuang, Mohammad Al Hasan
2022 arXiv   pre-print
successfully repair a GCN-based node classifier with superior performance than several competing methods.  ...  To overcome this issue, GraphLT subsequently uses a novel Bayesian label transition model, which takes GCN's predicted labels and applies label transitions by Gibbs-sampling-based inference and thus repairs  ...  For instance, we train GraphSAGE using the noisy labels with nr = 0.3 and get 69.67% original accuracy on Citeseer.  ... 
arXiv:2010.14121v3 fatcat:q3mgpp5ptjfxnll37gjhi5rywa

Defending Graph Convolutional Networks against Dynamic Graph Perturbations via Bayesian Self-supervision [article]

Jun Zhuang, Mohammad Al Hasan
2022 arXiv   pre-print
In this paper, we generalize noisy supervision as a kind of self-supervised learning method and then propose a novel Bayesian self-supervision model, namely GraphSS, to address the issue.  ...  Many existing works aim to strengthen the robustness of GCNs; for instance, adversarial training is used to shield GCNs against malicious perturbations.  ...  Acknowledgments This research is partially supported by National Science Foundation with grant number IIS-1909916.  ... 
arXiv:2203.03762v1 fatcat:b5oeyesb3bar3bsoiao3hzv2lq

Dense Structure Inference for Object Classification in Aerial LIDAR Dataset

Eunyoung Kim, Gerard Medioni
2010 2010 20th International Conference on Pattern Recognition  
To deal with sparse density in points representing each candidate, we also propose a novel method to infer a dense 3D structure from the given sparse and noisy points without any meshes and iterations.  ...  To label object candidates, we build a tree-structure database of object classes, which captures latent patterns in shape of 3D objects in a hierarchical manner.  ...  Large structure identification From the observation that the majority of large structures in the urban area are nearly planar, we can infer these structures by recognizing smoothly continuous planar patches  ... 
doi:10.1109/icpr.2010.747 dblp:conf/icpr/KimM10 fatcat:wbibv23ktrbzlkeynblsvuw4cu

Learning Graph-Structured Sum-Product Networks for Probabilistic Semantic Maps [article]

Kaiyu Zheng, Andrzej Pronobis, Rajesh P. N. Rao
2017 arXiv   pre-print
We introduce Graph-Structured Sum-Product Networks (GraphSPNs), a probabilistic approach to structured prediction for problems where dependencies between latent variables are expressed in terms of arbitrary  ...  While many approaches to structured prediction place strict constraints on the interactions between inferred variables, many real-world problems can be only characterized using complex graph structures  ...  In the future, by combining a Graph-SPN learned over semantic maps with the generative place model proposed in , we intend to achieve a unified, deep, and hierarchical representation of spatial knowledge  ... 
arXiv:1709.08274v2 fatcat:afjvtp7etzg4lingju3m6v5lde

Bag Graph: Multiple Instance Learning Using Bayesian Graph Neural Networks

Soumyasundar Pal, Antonios Valkanas, Florence Regol, Mark Coates
2022 PROCEEDINGS OF THE THIRTIETH AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE AND THE TWENTY-EIGHTH INNOVATIVE APPLICATIONS OF ARTIFICIAL INTELLIGENCE CONFERENCE  
Instead of focusing on each instance, these models are trained in an end-to-end fashion to learn effective bag-level representations by suitably combining permutation invariant pooling techniques with  ...  Multiple Instance Learning (MIL) is a weakly supervised learning problem where the aim is to assign labels to sets or bags of instances, as opposed to traditional supervised learning where each instance  ...  For the proposed Bayesian approach, we have two hyperparameters k and r associated with the approximate graph inference technique in (Kalofolias and Perraudin 2019) , used in Step 4 of Algorithm 1.  ... 
doi:10.1609/aaai.v36i7.20762 fatcat:kjtcol5yjne4zp4odbksawkwvu

Learning Graph-Structured Sum-Product Networks for Probabilistic Semantic Maps

Kaiyu Zheng, Andrzej Pronobis, Rajesh Rao
2018 PROCEEDINGS OF THE THIRTIETH AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE AND THE TWENTY-EIGHTH INNOVATIVE APPLICATIONS OF ARTIFICIAL INTELLIGENCE CONFERENCE  
We introduce Graph-Structured Sum-Product Networks (GraphSPNs), a probabilistic approach to structured prediction for problems where dependencies between latent variables are expressed in terms of arbitrary  ...  While many approaches to structured prediction place strict constraints on the interactions between inferred variables, many real-world problems can be only characterized using complex graph structures  ...  become invaluable in real-world settings where inference is performed with noisy data and under partial observability.  ... 
doi:10.1609/aaai.v32i1.11743 fatcat:o5lcrfnpmvdnlp7ysu6kcid5q4

Reconstructing Nonlinear Dynamical Systems from Multi-Modal Time Series [article]

Daniel Kramer, Philine Lou Bommer, Carlo Tombolini, Georgia Koppe, Daniel Durstewitz
2022 arXiv   pre-print
In many areas of science it is common to sample time series observations from many data modalities simultaneously, e.g. electrophysiological and behavioral time series in a typical neuroscience experiment  ...  Empirically observed time series in physics, biology, or medicine, are commonly generated by some underlying dynamical system (DS) which is the target of scientific interest.  ...  Improving DS reconstruction with multi-modal data when continuous observations are too noisy. A) Experimental setup with Gaussian and categorical information.  ... 
arXiv:2111.02922v3 fatcat:b24xlggdojgkfju5cnl55m47jy

A Goal-Directed Bayesian Framework for Categorization

Francesco Rigoli, Giovanni Pezzulo, Raymond Dolan, Karl Friston
2017 Frontiers in Psychology  
Assuming the brain performs Bayesian inference, based on a generative model of the external world and future goals, we propose a computational model of categorization in which important properties emerge  ...  Our account is a step toward elucidating computational principles of categorization and its role within the Bayesian brain hypothesis.  ...  In short, optimality in approximate Bayesian inference is a nuanced optimality that probably accounts for the intractable nature of exact Bayesian inference.  ... 
doi:10.3389/fpsyg.2017.00408 pmid:28382008 pmcid:PMC5360703 fatcat:7y6ak6yit5eohj4jrk7r7oxjba

Challenges in Markov chain Monte Carlo for Bayesian neural networks [article]

Theodore Papamarkou and Jacob Hinkle and M. Todd Young and David Womble
2021 arXiv   pre-print
Markov chain Monte Carlo (MCMC) methods have not been broadly adopted in Bayesian neural networks (BNNs).  ...  This paper initially reviews the main challenges in sampling from the parameter posterior of a neural network via MCMC. Such challenges culminate to lack of convergence to the parameter posterior.  ...  So, one of the Bayesian inference strategies for neural networks is to approximate the covariance structure among network parameters.  ... 
arXiv:1910.06539v6 fatcat:a7yyjtpsxvcd5okxwclt5gm3xe

A Nonparametric Model for Multimodal Collaborative Activities Summarization [article]

Guy Rosman, John W. Fisher III, Daniela Rus
2017 arXiv   pre-print
We further demonstrate how spatio-temporal structure embedded in our model enables better understanding of partial and noisy observations such as localization and face detections based on social interactions  ...  This is especially evident in urban environments teeming with human activities, but which suffer from incomplete and noisy data.  ...  By treating collaborative activity instances as latent factors with a spatiotemporal structure, we can summarize activities for co-located individuals despite noisy and partial/missing signals by pooling  ... 
arXiv:1709.01077v1 fatcat:33xdjwo3ybdipeezuifxq62eqm

Capturing Actionable Dynamics with Structured Latent Ordinary Differential Equations [article]

Paidamoyo Chapfuwa, Sherri Rose, Lawrence Carin, Edward Meeds, Ricardo Henao
2022 arXiv   pre-print
Results on challenging biological datasets show consistent improvements over competitive baselines in the controlled generation of observational data and inference of biologically meaningful system inputs  ...  To that end, we propose a structured latent ODE model that explicitly captures system input variations within its latent representation.  ...  In contrast, our structured modeling approach has significant benefits over baseline methods in model fit (or ELBO), due to its system input inference (9) and structured conditional prior (2).  ... 
arXiv:2202.12932v2 fatcat:s7s5o2g3yveclmx2gxe7yjeeca

Inferring Missing Categorical Information in Noisy and Sparse Web Markup

Nicolas Tempelmeier, Elena Demidova, Stefan Dietze
2018 Proceedings of the 2018 World Wide Web Conference on World Wide Web - WWW '18  
In this work, we introduce a supervised approach for inferring missing categorical properties in Web markup.  ...  For instance, from 26 million nodes describing events within the Common Crawl in 2016, 59% of nodes provide less than six statements and only 257,000 nodes (0.96%) are typed with more specific event subtypes  ...  Moreover, we observed that it is not sufficient to consider only node-specific features such as node-vocab to infer missing categorical information in Web markup.  ... 
doi:10.1145/3178876.3186028 dblp:conf/www/TempelmeierDD18 fatcat:xtawwkhokna5xb5swua63dztlu

Using Noisy Extractions to Discover Causal Knowledge [article]

Dhanya Sridhar, Jay Pujara, Lise Getoor
2017 arXiv   pre-print
We introduce a probabilistic method for fusing noisy extractions with observational data to discover causal knowledge.  ...  One approach attempts to identify causal relationships from text using automatic extraction techniques, while the other approach infers causation from observational data.  ...  However, in many problems, empirical observations of entities, or observational data, are readily available, potentially recovering information when fused with noisy extractions.  ... 
arXiv:1711.05900v1 fatcat:c6ele5eq4nf5zj65khc7xzmfqm

SAG-VAE: End-to-end Joint Inference of Data Representations and Feature Relations [article]

Chen Wang, Chengyuan Deng, Vladimir Ivanov
2020 arXiv   pre-print
Variational Autoencoders (VAEs) are powerful in data representation inference, but it cannot learn relations between features with its vanilla form and common variations.  ...  Experiments based on graphs show that SAG-VAE is capable of approximately retrieving edges and links between nodes based entirely on feature observations.  ...  Following the doctrine of variational inference, we use the Gumbel-Softmax distribution to approximate the probability for each edge: q φ (A s,t |X) = Gumbel-Softmax(φ 1 (A s,t |X)) (6) Notice that in  ... 
arXiv:1911.11984v3 fatcat:vejc2if3rffu3ihgtmz65ceybi
« Previous Showing results 1 — 15 out of 19,586 results