| 1. |
- Dang, Khue-Dung, et al.
(författare)
-
Hamiltonian Monte Carlo with Energy Conserving Subsampling
- 2019
-
Ingår i: Journal of machine learning research. - : MIT Press. - 1532-4435 .- 1533-7928. ; 20, s. 1-31
-
Tidskriftsartikel (refereegranskat)abstract
- Hamiltonian Monte Carlo (HMC) samples efficiently from high-dimensional posterior distributions with proposed parameter draws obtained by iterating on a discretized version of the Hamiltonian dynamics. The iterations make HMC computationally costly, especially in problems with large data sets, since it is necessary to compute posterior densities and their derivatives with respect to the parameters. Naively computing the Hamiltonian dynamics on a subset of the data causes HMC to lose its key ability to generate distant parameter proposals with high acceptance probability. The key insight in our article is that efficient subsampling HMC for the parameters is possible if both the dynamics and the acceptance probability are computed from the same data subsample in each complete HMC iteration. We show that this is possible to do in a principled way in a HMC-within-Gibbs framework where the subsample is updated using a pseudo marginal MH step and the parameters are then updated using an HMC step, based on the current subsample. We show that our subsampling methods are fast and compare favorably to two popular sampling algorithms that use gradient estimates from data subsampling. We also explore the current limitations of subsampling HMC algorithms by varying the quality of the variance reducing control variates used in the estimators of the posterior density and its gradients.
|
|
| 2. |
- Quiroz, Matias, et al.
(författare)
-
The Block-Poisson Estimator for Optimally Tuned Exact Subsampling MCMC
- 2021
-
Ingår i: Journal of Computational And Graphical Statistics. - : Informa UK Limited. - 1061-8600 .- 1537-2715. ; 30:4, s. 877-888
-
Tidskriftsartikel (refereegranskat)abstract
- Speeding up Markov chain Monte Carlo (MCMC) for datasets with many observations by data subsampling has recently received considerable attention. A pseudo-marginal MCMC method is proposed that estimates the likelihood by data subsampling using a block-Poisson estimator. The estimator is a product of Poisson estimators, allowing us to update a single block of subsample indicators in each MCMC iteration so that a desired correlation is achieved between the logs of successive likelihood estimates. This is important since pseudo-marginal MCMC with positively correlated likelihood estimates can use substantially smaller subsamples without adversely affecting the sampling efficiency. The block-Poisson estimator is unbiased but not necessarily positive, so the algorithm runs the MCMC on the absolute value of the likelihood estimator and uses an importance sampling correction to obtain consistent estimates of the posterior mean of any function of the parameters. Our article derives guidelines to select the optimal tuning parameters for our method and shows that it compares very favorably to regular MCMC without subsampling, and to two other recently proposed exact subsampling approaches in the literature. Supplementary materials for this article are available online.
|
|
| 3. |
- Tobías, Aurelio, et al.
(författare)
-
Geographical Variations of the Minimum Mortality Temperature at a Global Scale : A Multicountry Study
- 2021
-
Ingår i: Environmental epidemiology. - : Wolters Kluwer. - 2474-7882. ; 5:5
-
Tidskriftsartikel (refereegranskat)abstract
- Background: Minimum mortality temperature (MMT) is an important indicator to assess the temperature-mortality association, indicating long-term adaptation to local climate. Limited evidence about the geographical variability of the MMT is available at a global scale.Methods: We collected data from 658 communities in 43 countries under different climates. We estimated temperature-mortality associations to derive the MMT for each community using Poisson regression with distributed lag nonlinear models. We investigated the variation in MMT by climatic zone using a mixed-effects meta-analysis and explored the association with climatic and socioeconomic indicators.Results: The geographical distribution of MMTs varied considerably by country between 14.2 and 31.1 °C decreasing by latitude. For climatic zones, the MMTs increased from alpine (13.0 °C) to continental (19.3 °C), temperate (21.7 °C), arid (24.5 °C), and tropical (26.5 °C). The MMT percentiles (MMTPs) corresponding to the MMTs decreased from temperate (79.5th) to continental (75.4th), arid (68.0th), tropical (58.5th), and alpine (41.4th). The MMTs indreased by 0.8 °C for a 1 °C rise in a community's annual mean temperature, and by 1 °C for a 1 °C rise in its SD. While the MMTP decreased by 0.3 centile points for a 1 °C rise in a community's annual mean temperature and by 1.3 for a 1 °C rise in its SD.Conclusions: The geographical distribution of the MMTs and MMTPs is driven mainly by the mean annual temperature, which seems to be a valuable indicator of overall adaptation across populations. Our results suggest that populations have adapted to the average temperature, although there is still more room for adaptation.
|
|
| 4. |
- Kasiuk, Julia, et al.
(författare)
-
The enhancement of low-temperature excitation of magnons via interlayer exchange coupling in perpendicularly magnetized [Co/Pd] multilayers
- 2024
-
Ingår i: APPLIED PHYSICS LETTERS. - 0003-6951 .- 1077-3118. ; 124:19
-
Tidskriftsartikel (refereegranskat)abstract
- In this study, we analyze the correlation between magnetization and magnetoresistance of perpendicularly anisotropic [Co/Pd] multilayered films with different thicknesses of Pd layers t(Pd) = 0.6-2.0 nm in a wide range of temperatures, T = 4-300 K. We revealed that electron scattering by magnons makes a significant contribution to the magnetoresistance of the multilayers regardless of the layer thickness. Contrary to expectations, the effect of magnon magnetoresistance (MMR) increases with decreasing temperature below T = 50 K in the films with t(Pd) = 0.8 and 1.0 nm. The revealed low-temperature MMR increase, which is most pronounced in the [Co-0.5/Pd-1.0] multilayers, is associated with the enhanced magnon excitation due to antiferromagnetic exchange coupling between the Co layers. The latter ensures an atypical shape of the magnetization curves of the [Co-0.5/Pd-1.0] multilayers at low temperatures in a perpendicular magnetic field, which combine a quadratic hysteresis loop of a perpendicularly anisotropic ferromagnet and an anomalous magnetization drop resulting from a violation of the ordering of magnetic moments and their amplified oscillations initiated by the interlayer exchange coupling.
|
|
| 5. |
- Gasparrini, Antonio, et al.
(författare)
-
Projections of temperature-related excess mortality under climate change scenarios
- 2017
-
Ingår i: The Lancet Planetary Health. - 2542-5196. ; 1:9, s. e360-e367
-
Tidskriftsartikel (refereegranskat)abstract
- Background: Climate change can directly affect human health by varying exposure to non-optimal outdoor temperature. However, evidence on this direct impact at a global scale is limited, mainly due to issues in modelling and projecting complex and highly heterogeneous epidemiological relationships across different populations and climates.Methods: We collected observed daily time series of mean temperature and mortality counts for all causes or non-external causes only, in periods ranging from Jan 1, 1984, to Dec 31, 2015, from various locations across the globe through the Multi-Country Multi-City Collaborative Research Network. We estimated temperature-mortality relationships through a two-stage time series design. We generated current and future daily mean temperature series under four scenarios of climate change, determined by varying trajectories of greenhouse gas emissions, using five general circulation models. We projected excess mortality for cold and heat and their net change in 1990-2099 under each scenario of climate change, assuming no adaptation or population changes.Findings: Our dataset comprised 451 locations in 23 countries across nine regions of the world, including 85 879 895 deaths. Results indicate, on average, a net increase in temperature-related excess mortality under high-emission scenarios, although with important geographical differences. In temperate areas such as northern Europe, east Asia, and Australia, the less intense warming and large decrease in cold-related excess would induce a null or marginally negative net effect, with the net change in 2090-99 compared with 2010-19 ranging from -1·2% (empirical 95% CI -3·6 to 1·4) in Australia to -0·1% (-2·1 to 1·6) in east Asia under the highest emission scenario, although the decreasing trends would reverse during the course of the century. Conversely, warmer regions, such as the central and southern parts of America or Europe, and especially southeast Asia, would experience a sharp surge in heat-related impacts and extremely large net increases, with the net change at the end of the century ranging from 3·0% (-3·0 to 9·3) in Central America to 12·7% (-4·7 to 28·1) in southeast Asia under the highest emission scenario. Most of the health effects directly due to temperature increase could be avoided under scenarios involving mitigation strategies to limit emissions and further warming of the planet.Interpretation: This study shows the negative health impacts of climate change that, under high-emission scenarios, would disproportionately affect warmer and poorer regions of the world. Comparison with lower emission scenarios emphasises the importance of mitigation policies for limiting global warming and reducing the associated health risks.
|
|
| 6. |
- Quiroz, Matias, et al.
(författare)
-
Subsampling MCMC - an Introduction for the Survey Statistician
- 2018
-
Ingår i: SANKHYA-SERIES A-MATHEMATICAL STATISTICS AND PROBABILITY. - : SPRINGER. - 0976-836X. ; 80, s. 33-69
-
Forskningsöversikt (refereegranskat)abstract
- The rapid development of computing power and efficient Markov Chain Monte Carlo (MCMC) simulation algorithms have revolutionized Bayesian statistics, making it a highly practical inference method in applied work. However, MCMC algorithms tend to be computationally demanding, and are particularly slow for large datasets. Data subsampling has recently been suggested as a way to make MCMC methods scalable on massively large data, utilizing efficient sampling schemes and estimators from the survey sampling literature. These developments tend to be unknown by many survey statisticians who traditionally work with non-Bayesian methods, and rarely use MCMC. Our article explains the idea of data subsampling in MCMC by reviewing one strand of work, Subsampling MCMC, a so called Pseudo-Marginal MCMC approach to speeding up MCMC through data subsampling. The review is written for a survey statistician without previous knowledge of MCMC methods since our aim is to motivate survey sampling experts to contribute to the growing Subsampling MCMC literature.
|
|
| 7. |
- Wu, Wen-Bin, et al.
(författare)
-
Observation of higher-order contribution to anisotropic magnetoresistance of thin Pt/[Co/Pt] multilayered films
- 2024
-
Ingår i: APPLIED SURFACE SCIENCE. - 0169-4332 .- 1873-5584. ; 648
-
Tidskriftsartikel (refereegranskat)abstract
- We studied the magnetoresistance mechanisms in a Pt/[Co/Pt]x5 film consisting of a ferromagnetic [Co/Pt]x5 layer with strong perpendicular magnetic anisotropy and a nonmagnetic Pt layer with strong spin-orbit coupling. We revealed two competing contributions of the sin2 theta and cos4 theta types to its angular and magnetic field de-pendences of electrical resistance at T = 10-250 K corresponding to the out-of-plane rotation of the magneti-zation M(theta) perpendicularly to the electric current. They were attributed to different magnetoresistance mechanisms. The higher-order cos4 theta contribution, which emerges and increases with decreasing temperature, is attributed to the anisotropic magnetoresistance of the ferromagnetic layer, while the sin2 theta contribution, which prevails at room temperature and then decreases, is mainly associated with the spin Hall magnetoresistance originating from the Pt layer. The analysis of the corresponding angular dependences of the Hall voltage revealed non-trivial periodic oscillations in the second harmonic. Their appearance is found to be consistent with the manifestation of higher-order angle-dependent contributions to the field-like spin-orbit torque. The revealed strong influence of the electric current on the magnetization of the film studied, which ensures the higher-order effects manifestation, is of high relevance for magnetic memory design technologies.
|
|
