scispace - formally typeset
Search or ask a question
Topic

Latent variable model

About: Latent variable model is a research topic. Over the lifetime, 3589 publications have been published within this topic receiving 235061 citations.


Papers
More filters
Posted Content
TL;DR: This tutorial explores issues in depth through the lens of variational inference about how to parameterize conditional likelihoods in latent variable models with powerful function approximators.
Abstract: There has been much recent, exciting work on combining the complementary strengths of latent variable models and deep learning. Latent variable modeling makes it easy to explicitly specify model constraints through conditional independence properties, while deep learning makes it possible to parameterize these conditional likelihoods with powerful function approximators. While these "deep latent variable" models provide a rich, flexible framework for modeling many real-world phenomena, difficulties exist: deep parameterizations of conditional likelihoods usually make posterior inference intractable, and latent variable objectives often complicate backpropagation by introducing points of non-differentiability. This tutorial explores these issues in depth through the lens of variational inference.

41 citations

Journal ArticleDOI
TL;DR: In this paper, an integrated choice and latent variable model is used to explicitly measure the perceived advantages in electric vehicles over the conventional internal combustion engine vehicles, and the results show higher probability of adopting electric vehicles for Gen Y, compared to Gen X and after them.
Abstract: Relative advantage, or the degree to which a new technology is perceived to be better than an existing technology which is being replaced, has a significant impact on individuals’ decisions on when, how and to what extent to adopt. An integrated choice and latent variable model is used, in this paper, to explicitly measure the perceived advantages in electric vehicles over the conventional internal combustion engine vehicles. The analysed data is obtained from a stated preference survey including 1076 residents in New South Wales, Australia. According to the results, the latent component of the model disentangles the perceived advantages across three dimensions of vehicle design, impact on the environment, and safety. These latent variables are interacted with price, driving range and body type, respectively, to capture the impact of perception on preference. The developed model is then used to examine the effectiveness of different support schemes on Millennials (Gen Y), the generation before them (Gen X) and after them (Gen Z). The results show higher probability of adopting electric vehicles for Gen Y, compared to Gen X and Z. Gen Y is found to be the least sensitive cohort to purchase price, and Gen X to be the most sensitive cohort to this attribute. People are more sensitive to incentives for the initial price compared to ongoing incentives for operating costs. Also, offering financial incentives to consumers as a rebate on the purchase price is more effective than allocating the same incentive to manufacturers to reduce the purchase price.

40 citations

Book ChapterDOI
08 Nov 2010
TL;DR: This work learns latent spaces, and distributions within them, for image features and 3D poses separately first, and then learns a multi-modal conditional density between these two lowdimensional spaces in the form of Gaussian Mixture Regression.
Abstract: Discriminative approaches for human pose estimation model the functional mapping, or conditional distribution, between image features and 3D pose. Learning such multi-modal models in high dimensional spaces, however, is challenging with limited training data; often resulting in over-fitting and poor generalization. To address these issues latent variable models (LVMs) have been introduced. Shared LVMs attempt to learn a coherent, typically non-linear, latent space shared by image features and 3D poses, distribution of data in that latent space, and conditional distributions to and from this latent space to carry out inference. Discovering the shared manifold structure can, in itself, however, be challenging. In addition, shared LVMs models are most often non-parametric, requiring the model representation to be a function of the training set size. We present a parametric framework that addresses these shortcoming. In particular, we learn latent spaces, and distributions within them, for image features and 3D poses separately first, and then learn a multi-modal conditional density between these two lowdimensional spaces in the form of Gaussian Mixture Regression. Using our model we can address the issue of over-fitting and generalization, since the data is denser in the learned latent space, as well as avoid the necessity of learning a shared manifold for the data. We quantitatively evaluate and compare the performance of the proposed method to several state-of-the-art alternatives, and show that our method gives a competitive performance.

40 citations

Journal ArticleDOI
12 Apr 2017-PLOS ONE
TL;DR: The findings suggest a complex, multidimensional mental health structure in the youth population rather than the previously assumed first or second order factor structure, and reveal a low hazardous behaviour/low mental illness risk subgroup not previously described.
Abstract: Little is known about the underlying relationships between self-reported mental health items measuring both positive and negative emotional and behavioural symptoms at the population level in young people. Improved measurement of the full range of mental well-being and mental illness may aid in understanding the aetiological substrates underlying the development of both mental wellness as well as specific psychiatric diagnoses. A general population sample aged 14 to 24 years completed self-report questionnaires on anxiety, depression, psychotic-like symptoms, obsessionality and well-being. Exploratory and confirmatory factor models for categorical data and latent profile analyses were used to evaluate the structure of both mental wellness and illness items. First order, second order and bifactor structures were evaluated on 118 self-reported items obtained from 2228 participants. A bifactor solution was the best fitting latent variable model with one general latent factor termed ‘distress’ and five ‘distress independent’ specific factors defined as self-confidence, antisocial behaviour, worry, aberrant thinking, and mood. Next, six distinct subgroups were derived from a person-centred latent profile analysis of the factor scores. Finally, concurrent validity was assessed using information on hazardous behaviours (alcohol use, substance misuse, self-harm) and treatment for mental ill health: both discriminated between the latent traits and latent profile subgroups. The findings suggest a complex, multidimensional mental health structure in the youth population rather than the previously assumed first or second order factor structure. Additionally, the analysis revealed a low hazardous behaviour/low mental illness risk subgroup not previously described. Population sub-groups show greater validity over single variable factors in revealing mental illness risks. In conclusion, our findings indicate that the structure of self reported mental health is multidimensional in nature and uniquely finds improved prediction to mental illness risk within person-centred subgroups derived from the multidimensional latent traits.

40 citations

Journal ArticleDOI
TL;DR: In general, for factorial designs, an analysis of variance of the observed variable Y cannot be used to draw inferences about main effects and interactions on the latent variable θ even when the standard normality and equality of variance assumptions hold as mentioned in this paper.
Abstract: Let Y be a continuous, ordinal measure of a latent variable θ. In general, for factorial designs, an analysis of variance of the observed variable Y cannot be used to draw inferences about main effects and interactions on the latent variable θ even when the standard normality and equality of variance assumptions hold.

40 citations


Network Information
Related Topics (5)
Statistical hypothesis testing
19.5K papers, 1M citations
82% related
Inference
36.8K papers, 1.3M citations
81% related
Multivariate statistics
18.4K papers, 1M citations
80% related
Linear model
19K papers, 1M citations
80% related
Estimator
97.3K papers, 2.6M citations
78% related
Performance
Metrics
No. of papers in the topic in previous years
YearPapers
202375
2022143
2021137
2020185
2019142
2018159