Bayesian sparse factor analysis with kernelized observations

Sevilla-Salcedo, Carlos; Guerrero-López, Alejandro; Olmos, Pablo M.; Gómez-Verdejo, Vanessa

doi:10.1016/j.neucom.2022.03.024

Cited by 3 publications

(14 citation statements)

References 22 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In this context, we propose to adapt KSSHIBA [34] model, the kernelized version of SSHIBA [35], to deal with MALDI-TOF MS data so that we can efficiently predict the CP and ESBL susceptibility for each isolate. KSSHIBA is a Bayesian multiview semi-supervised model intended to deal with high-dimensional data because, on the one hand, works in the data space ( N × N ) being N << D by means of using kernel data representations, and, on the other hand, it projects all input views to a common low-dimensional latent space.…”

Section: Methodsmentioning

confidence: 99%

See 1 more Smart Citation

Automatic antibiotic resistance prediction inKlebsiella pneumoniaebased on MALDI-TOF mass spectra

Guerrero-López

Candela

Sevilla-Salcedo

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

Objectives: MALDI-TOF Mass Spectrometry (MS) is a reference method for microbial identification at clinical microbiology laboratories. We have designed and validated a new multiview model based on machine learning from MS spectra to predict antibiotic resistance mechanisms 24 h before phenotypic results are available. Methods: Antibiotic susceptibility of 402 clinical Klebsiella pneumoniae isolates was determined in two collections, discriminating among Wild Type (WT), Extended-Spectrum Beta-Lactamases (ESBL) producers, and ESBL and Carbapenemases (ESBL+CP) producers. Each isolate was subcultured 3 consecutive days and 2 independent spectra were acquired in each replica (6 MS spectra/isolate). Spectra were automatically classified by a kernelized Bayesian factor analysis model (KSSHIBA), using two independent strategies: 1) the model was designed with isolates from a single center and validated with isolates from the other center; and 2) in a second stage all isolates were used at the same time for design and validation processes. Results: Higher prediction values were obtained when integrating all isolates with hospital collection of origin information. Our model exhibited higher prediction capability than current state-of-the-art models, particularly in intercollection scenarios because local epidemiology could introduce relevant variables affecting prediction accuracy. Conclusions: Compared to previously reported studies, our model demonstrated the highest ability to predict ESBL and/or CP production in clinical K. pneumoniae isolates and it provided an efficient way to combine information from different centers. Its implementation in microbiological laboratories could improve the detection of multi-drug resistant isolates, optimizing the therapeutic decision.

show abstract

Section: Methodsmentioning

confidence: 99%

“…These posteriors are approximated through mean-field variational inference [47] maximising the Evidence Lower BOund (ELBO). For more details, see [43, 44]. Furthermore, the Bayesian nature of the model allows it to work in a semi-supervised fashion, using all available information to determine the approximate distribution of the variables.…”

Section: Methodsmentioning

confidence: 99%

Automatic antibiotic resistance prediction inKlebsiella pneumoniaebased on MALDI-TOF mass spectra

Guerrero-López

Candela

Sevilla-Salcedo

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…Once the background have been established, our first technical contribution is presented. In this chapter, we review the kernelised version of SSHIBA presented in [122]. First, we present the technical advances of KSSHIBA in terms of dealing with kernelised views, automatic Bayesian relevance vector selection, and automatic feature-relevance determination.…”

Section: Organisationmentioning

confidence: 99%

“…To address this limitation, this thesis presents various extensions to the SSHIBA model to better suit it for microbiological data. In Chapter 3, a kernel-based extension, KSSHIBA [122], is proposed to handle high-dimensional data, such as that obtained from MALDI-TOF MS, while exploiting non-linear relationships in the data. Then, in Chapter 4, a VAE-based extension, FA-VAE [123], is proposed to exploit non-linear relationships and further expand the types of heterogeneous data to handle.…”

Section: Sparse Semi-supervised Heterogeneous Interbattery Bayesian A...mentioning

confidence: 99%

Automatic antibiotic resistance prediction in Klebsiella pneumoniae based on MALDI-TOF mass spectra

Guerrero-López

Sevilla-Salcedo

Candela

et al. 2023

Engineering Applications of Artificial Intelligence

Self Cite

View full text Add to dashboard Cite

“…SSHIBA [26,27] presents a solution to multi-view problems with samples represented in M different modalities where each view can be either multilabel, binary, real, categorical, or other multidimensional object. The general model framework, depicted in Fig.…”

Section: Sparse Semi-supervised Heterogeneous Interbattery Bayesian A...mentioning

confidence: 99%

Multimodal hierarchical Variational AutoEncoders with Factor Analysis latent space

Guerrero-López¹,

Sevilla-Salcedo²,

Gómez-Verdejo³

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

Real-world databases are complex, they usually present redundancy and shared correlations between heterogeneous and multiple representations of the same data. Thus, exploiting and disentangling shared information between views is critical. For this purpose, recent studies often fuse all views into a shared nonlinear complex latent space but they lose the interpretability. To overcome this limitation, here we propose a novel method to combine multiple Variational AutoEncoders (VAE) architectures with a Factor Analysis latent space (FA-VAE).Concretely, we use a VAE to learn a private representation of each heterogeneous view in a continuous latent space. Then, we model the shared latent space by projecting every private variable to a low-dimensional latent space using a linear projection matrix. Thus, we create an interpretable hierarchical dependency between private and shared information. This way, the novel model is able to simultaneously: (i) learn from multiple heterogeneous views, (ii) obtain an interpretable hierarchical shared space, and, (iii) perform transfer learning between generative models.

show abstract

Bayesian sparse factor analysis with kernelized observations

Cited by 3 publications

References 22 publications

Automatic antibiotic resistance prediction inKlebsiella pneumoniaebased on MALDI-TOF mass spectra

Automatic antibiotic resistance prediction inKlebsiella pneumoniaebased on MALDI-TOF mass spectra

Automatic antibiotic resistance prediction in Klebsiella pneumoniae based on MALDI-TOF mass spectra

Multimodal hierarchical Variational AutoEncoders with Factor Analysis latent space

Contact Info

Product

Resources

About