Variational learning for rectified factor analysis

Harva, Markus; Kabán, Ata

doi:10.1016/j.sigpro.2006.06.006

Cited by 29 publications

(21 citation statements)

References 14 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Each array of d for a difference frame provides a short-term (i.e., frame to frame) measure of dispersion, therefore we elected to use the standard deviation of d (sd) for each difference frame (i.e., standard deviation of the 307,200 data points per difference frame). This choice was based on the fact that rectifying a gaussian distribution more strongly affects the mean (µ) than the variance (s 2 ), and thus selecting the standard deviation (s) of the difference image data would not magnify gaussian noise as much, and thereby enhance the signal-to-noise ratio (Harva and Kaban 2007). A plot of the standard deviation of di vs. time (Figure 2a) provides a signal that relates to short-term variation.…”

Section: Image Analysis Processmentioning

confidence: 99%

“…More recently, portable and non-invasive 3D accelerometry has been employed to measure motion and activity; numerous studies have calibrated or assessed the ability of motionbased sensors with resultant g forces to adequately assess metabolic expenditure (Gleiss et al 2011;Wright et al 2014;Hicks et al 2017;Jeanniard-du-Dot et al 2017). Floor-switch sensors have been used in laboratory situations to estimate movement (Spiteri 1982;Harva and Kaban 2007), in the same manner that beam crossing active infrared approaches have been employed in studies of sleep in animal activity (Parrish and Teske 2017). Numerous analytical approaches to electrical representations of electromyography exist, including root mean square analysis, zero crossing detection, linear envelope, time integration, and power spectrum analysis (Kwatny et al 1970;Siegler et al 1985;Nilsson et al 1993;Winter and Patla 1997).…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Activity analysis of thermal imaging videos using a difference imaging approach

Tattersall

Danner

Chaves

et al. 2020

Journal of Thermal Biology

View full text Add to dashboard Cite

Infrared thermal imaging is a passive imaging technique that captures the emitted radiation from an object to estimate surface temperature, often for inference of heat transfer. Infrared thermal imaging offers the potential to detect movement without the challenges of glare, shadows, or changes in lighting associated with visual digital imaging or active infrared imaging. In this paper, we employ a frame subtraction algorithm for extracting the pixel-by-pixel relative change in signal from a fixed focus video file, tailored for use with thermal imaging videos. By summing the absolute differences across an entire video, we are able to assign quantitative activity assessments to thermal imaging data for comparison with simultaneous recordings of metabolic rates. We tested the accuracy and limits of this approach by analyzing movement of a metronome and provide an example application of the approach to a study of Darwin's finches. In principle, this "Difference Imaging Thermography" (DIT) would allow for activity data to be standardized to energetic measurements and could be applied to any radiometric imaging system.

show abstract

Section: Image Analysis Processmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Activity analysis of thermal imaging videos using a difference imaging approach

Tattersall

Danner

Chaves

et al. 2020

Journal of Thermal Biology

View full text Add to dashboard Cite

show abstract

“…As noted in previous works [55], [56], closed form inference computations using a multivariate RG distribution are tractable only if the location parameter is zero (by effectively getting rid of the erfc(.) term).…”

Section: Rectified Gaussian Scale Mixturesmentioning

confidence: 99%

Rectified Gaussian Scale Mixtures and the Sparse Non-Negative Least Squares Problem

Nalci

Fedorov

Al-Shoukairi

et al. 2018

IEEE Trans. Signal Process.

View full text Add to dashboard Cite

In this paper, we develop a Bayesian evidence maximization framework to solve the sparse non-negative least squares (S-NNLS) problem. We introduce a family of probability densities referred to as the Rectified Gaussian Scale Mixture (R-GSM) to model the sparsity enforcing prior distribution for the solution. The R-GSM prior encompasses a variety of heavy-tailed densities such as the rectified Laplacian and rectified Studentt distributions with a proper choice of the mixing density. We utilize the hierarchical representation induced by the R-GSM prior and develop an evidence maximization framework based on the Expectation-Maximization (EM) algorithm. Using the EM based method, we estimate the hyper-parameters and obtain a point estimate for the solution. We refer to the proposed method as rectified sparse Bayesian learning (R-SBL). We provide four R-SBL variants that offer a range of options for computational complexity and the quality of the E-step computation. These methods include the Markov chain Monte Carlo EM, linear minimum mean-square-error estimation, approximate message passing and a diagonal approximation. Using numerical experiments, we show that the proposed R-SBL method outperforms existing S-NNLS solvers in terms of both signal and support recovery performance, and is also very robust against the structure of the design matrix.

show abstract

“…Considering the zero mean of the rectified Gaussian distribution [29], i.e., setū τ = 0, as approximated as…”

Section: B Proposed Variable Regularized Is-vrnmf2dmentioning

confidence: 99%

Machine Learning Source Separation Using Maximum a Posteriori Nonnegative Matrix Factorization

Gao

Woo

Ling

2014

IEEE Trans. Cybern.

View full text Add to dashboard Cite

A novel unsupervised machine learning algorithm for single channel source separation is presented. The proposed method is based on nonnegative matrix factorization, which is optimized under the framework of maximum a posteriori probability and Itakura-Saito divergence. The method enables a generalized criterion for variable sparseness to be imposed onto the solution and prior information to be explicitly incorporated through the basis vectors. In addition, the method is scale invariant where both low and high energy components of a signal are treated with equal importance. The proposed algorithm is a more complete and efficient approach for matrix factorization of signals that exhibit temporal dependency of the frequency patterns. Experimental tests have been conducted and compared with other algorithms to verify the efficiency of the proposed method.

show abstract

Variational learning for rectified factor analysis

Cited by 29 publications

References 14 publications

Activity analysis of thermal imaging videos using a difference imaging approach

Activity analysis of thermal imaging videos using a difference imaging approach

Rectified Gaussian Scale Mixtures and the Sparse Non-Negative Least Squares Problem

Machine Learning Source Separation Using Maximum a Posteriori Nonnegative Matrix Factorization

Contact Info

Product

Resources

About