scholarly journals Data-Dependent Conditional Priors for Unsupervised Learning of Multimodal Data

Entropy ◽  
2020 ◽  
Vol 22 (8) ◽  
pp. 888
Author(s):  
Frantzeska Lavda ◽  
Magda Gregorová ◽  
Alexandros Kalousis

One of the major shortcomings of variational autoencoders is the inability to produce generations from the individual modalities of data originating from mixture distributions. This is primarily due to the use of a simple isotropic Gaussian as the prior for the latent code in the ancestral sampling procedure for data generations. In this paper, we propose a novel formulation of variational autoencoders, conditional prior VAE (CP-VAE), with a two-level generative process for the observed data where continuous z and a discrete c variables are introduced in addition to the observed variables x. By learning data-dependent conditional priors, the new variational objective naturally encourages a better match between the posterior and prior conditionals, and the learning of the latent categories encoding the major source of variation of the original data in an unsupervised manner. Through sampling continuous latent code from the data-dependent conditional priors, we are able to generate new samples from the individual mixture components corresponding, to the multimodal structure over the original data. Moreover, we unify and analyse our objective under different independence assumptions for the joint distribution of the continuous and discrete latent variables. We provide an empirical evaluation on one synthetic dataset and three image datasets, FashionMNIST, MNIST, and Omniglot, illustrating the generative performance of our new model comparing to multiple baselines.


Author(s):  
S. S. Kramarenko ◽  
N. I. Kuzmichova ◽  
A. S. Kramarenko

The analysis included data on the origin and milk productivity of 109 first-born red steppe breed, which were descendants of five bulls-offspring (Narcissus, Topol, Tangens, Neptune, and Orpheus) and were kept in SE “Plemproductor Stepove” (Mykolaiv region, Ukraine ) during the years 2001–2014. The purpose of this study was to analyze the fat content of milk during different months of lactation (MFP1, MFP2,…, MFP10) to determine latent variables that best describe the variability of dairy cows' productivity in this herd. High correlation estimates of fat milk scores in different lactation months have been established. According to the results of the Principal Component Analysis, based on the (co)variation matrix of fat content in milk, three new variables (PC1, PC2, and PC3) were identified, which accounted for about 82% of the variability of the original data. The First Main Component (PC1) explained 53.5%, Second (PC2) – 17.7%, and Third (PC3) – 10.6% of the variability of the original data, respectively. PC1 was highly correlated with MFP4-MFP10 and, thus, it distributed the animals according to their fat content level. PC2 was highly positively correlated with MFP8-MFP10 but highly negatively correlated with M FP1-MFP3 and thus it shows the rate of increase in fat content in milk during lactation. PC3 characterizes the variability of fat content in milk during the first and second half of lactation. The Linear Discriminant Analysis found that the MFP1-MFP2 and MFP9-MFP10 scores contributed most to the discrimination among the five subpopulations. The individual identification of the offspring groups of different bulls according to the cross-check classification ranged from 44.4% (Topol) to 87.5% (Orpheus) of cows, which were correctly assigned to their own group.



2021 ◽  
pp. 016173462199809
Author(s):  
Dhurgham Al-karawi ◽  
Hisham Al-Assam ◽  
Hongbo Du ◽  
Ahmad Sayasneh ◽  
Chiara Landolfo ◽  
...  

Significant successes in machine learning approaches to image analysis for various applications have energized strong interest in automated diagnostic support systems for medical images. The evolving in-depth understanding of the way carcinogenesis changes the texture of cellular networks of a mass/tumor has been informing such diagnostics systems with use of more suitable image texture features and their extraction methods. Several texture features have been recently applied in discriminating malignant and benign ovarian masses by analysing B-mode images from ultrasound scan of the ovary with different levels of performance. However, comparative performance evaluation of these reported features using common sets of clinically approved images is lacking. This paper presents an empirical evaluation of seven commonly used texture features (histograms, moments of histogram, local binary patterns [256-bin and 59-bin], histograms of oriented gradients, fractal dimensions, and Gabor filter), using a collection of 242 ultrasound scan images of ovarian masses of various pathological characteristics. The evaluation examines not only the effectiveness of classification schemes based on the individual texture features but also the effectiveness of various combinations of these schemes using the simple majority-rule decision level fusion. Trained support vector machine classifiers on the individual texture features without any specific pre-processing, achieve levels of accuracy between 75% and 85% where the seven moments and the 256-bin LBP are at the lower end while the Gabor filter is at the upper end. Combining the classification results of the top k ( k = 3, 5, 7) best performing features further improve the overall accuracy to a level between 86% and 90%. These evaluation results demonstrate that each of the investigated image-based texture features provides informative support in distinguishing benign or malignant ovarian masses.



2004 ◽  
Vol 34 (1) ◽  
pp. 37-52
Author(s):  
Wiktor Jassem ◽  
Waldemar Grygiel

The mid-frequencies and bandwidths of formants 1–5 were measured at targets, at plus 0.01 s and at minus 0.01 s off the targets of vowels in a 100-word list read by five male and five female speakers, for a total of 3390 10-variable spectrum specifications. Each of the six Polish vowel phonemes was represented approximately the same number of times. The 3390* 10 original-data matrix was processed by probabilistic neural networks to produce a classification of the spectra with respect to (a) vowel phoneme, (b) identity of the speaker, and (c) speaker gender. For (a) and (b), networks with added input information from another independent variable were also used, as well as matrices of the numerical data appropriately normalized. Mean scores for classification with respect to phonemes in a multi-speaker design in the testing sets were around 95%, and mean speaker-dependent scores for the phonemes varied between 86% and 100%, with two speakers scoring 100% correct. The individual voices were identified between 95% and 96% of the time, and classifications of the spectra for speaker gender were practically 100% correct.



2021 ◽  
Vol 20 (5s) ◽  
pp. 1-25
Author(s):  
Shounak Chakraborty ◽  
Sangeet Saha ◽  
Magnus Själander ◽  
Klaus Mcdonald-Maier

Achieving high result-accuracy in approximate computing (AC) based real-time applications without violating power constraints of the underlying hardware is a challenging problem. Execution of such AC real-time tasks can be divided into the execution of the mandatory part to obtain a result of acceptable quality, followed by a partial/complete execution of the optional part to improve accuracy of the initially obtained result within the given time-limit. However, enhancing result-accuracy at the cost of increased execution length might lead to deadline violations with higher energy usage. We propose Prepare , a novel hybrid offline-online approximate real-time task-scheduling approach, that first schedules AC-based tasks and determines operational processing speeds for each individual task constrained by system-wide power limit, deadline, and task-dependency. At runtime, by employing fine-grained DVFS, the energy-adaptive processing speed governing mechanism of Prepare reduces processing speed during each last level cache miss induced stall and scales up the processing speed once the stall finishes to a higher value than the predetermined one. To ensure on-chip thermal safety, this higher processing speed is maintained only for a short time-span after each stall, however, this reduces execution times of the individual task and generates slacks. Prepare exploits the slacks either to enhance result-accuracy of the tasks, or to improve thermal and energy efficiency of the underlying hardware, or both. With a 70 - 80% workload, Prepare offers 75% result-accuracy with its constrained scheduling, which is enhanced by 5.3% for our benchmark based evaluation of the online energy-adaptive mechanism on a 4-core based homogeneous chip multi-processor, while meeting the deadline constraint. Overall, while maintaining runtime thermal safety, Prepare reduces peak temperature by up to 8.6 °C for our baseline system. Our empirical evaluation shows that constrained scheduling of Prepare outperforms a state-of-the-art scheduling policy, whereas our runtime energy-adaptive mechanism surpasses two current DVFS based thermal management techniques.



Author(s):  
Yuta Ojima ◽  
Eita Nakamura ◽  
Katsutoshi Itoyama ◽  
Kazuyoshi Yoshii

This paper describes automatic music transcription with chord estimation for music audio signals. We focus on the fact that concurrent structures of musical notes such as chords form the basis of harmony and are considered for music composition. Since chords and musical notes are deeply linked with each other, we propose joint pitch and chord estimation based on a Bayesian hierarchical model that consists of an acoustic model representing the generative process of a spectrogram and a language model representing the generative process of a piano roll. The acoustic model is formulated as a variant of non-negative matrix factorization that has binary variables indicating a piano roll. The language model is formulated as a hidden Markov model that has chord labels as the latent variables and emits a piano roll. The sequential dependency of a piano roll can be represented in the language model. Both models are integrated through a piano roll in a hierarchical Bayesian manner. All the latent variables and parameters are estimated using Gibbs sampling. The experimental results showed the great potential of the proposed method for unified music transcription and grammar induction.



2020 ◽  
Author(s):  
Francisco Freitas ◽  
Mónica Alves

AbstractBackgroundGuidelines for venous blood sampling procedure (phlebotomy) discourage tourniquet use whenever possible. Here, we aimed to assess the Biomedical Scientists capability of not using the tourniquet in phlebotomy, which we hypothesized to be equal to 50% of the patients attended, and identifying the most frequent venipuncture site.Materials and MethodsWe selected and assigned two (BMS) with the same age (41 years) and experience (20 years) to record ten phlebotomy days, the first with prioritized and the latter with non-prioritized patients. In a simple record form, each acquired daily data for the number of attended patients, age and gender, the frequency of non-tourniquet usage and the punctured vein. To test our work hypothesis we used the two-tailed single sample t-test (p < 0.05). Differences between age-group means and non-tourniquet use means by each BMS were tested by two-tailed t-test for independent means (p < 0.05).ResultsIn 10 phlebotomy days 683 patients were attended, with males representing 43,2% of the population. We found no statistically difference between age-group means. The combined capability of non-tourniquet use was 50,5%, which did not differ from our null hypothesis, but the individual group-means were statistically different, being 33% and 66.9% in the prioritized vs non-prioritized group. The medial cubital vein was the most prone to be punctured (77,7%).ConclusionsWe have shown that performing phlebotomies without tourniquet use is possible and desirable in at least half of the attended patients, though being more limited in specific group populations. Our results provide room for quality improvement in the laboratory pre-analytical phase.Key points summaryWe assessed the capability of Biomedical Scientists not using the tourniquet in real life blood sampling procedures for diagnostic purposes.Blood was collected from at least half of the attended patients without tourniquet use.Biomedical Scientists were able to prioritize the antecubital veins without tourniquet application (medial cubital vein the most prone to be punctured - 78% of attempts).



10.28945/2886 ◽  
2005 ◽  
Author(s):  
Tuija Stutzle ◽  
Jorma Sajaniemi

Roles of variables, which describe stereotypic usages of variables, can be exploited to facilitate teaching introductory programming. This paper describes the evaluation of visual metaphors for roles used in a role-based program animator. The evaluation is based on several criteria: properties of the images, metaphor recognition and grading, and effects on learning. The study demonstrates that as a whole the role metaphors facilitate learning. The results also identify ideas for further elaboration of the individual metaphors. Furthermore, the study suggests that the evaluation of animated metaphors may require special measures.



Author(s):  
Sandeep Krishnamurthy

E-mail is a low-cost and highly effective form of individual contact for primary research. However, researchers who contact strangers for their survey research through e-mail are, in essence, sending them Spam. Some academic researchers might argue that due to the low volume and infrequent nature of their surveys and the general positive perception of academia, their e-mail surveys do not add to the Spam problem. However, this is an insufficient resolution of the ethical problem. This chapter examines one solution to avoid this problem—the use of respondent permission prior to contact. Obtaining respondent permission is tricky and can be costly. But, it may be the only long-term solution. Importantly, using this approach could lead to a loss of randomness in the sampling procedure due to self-selection. Ideas for implementation of a permission-based contact system at the individual researcher and academic field level are provided at the end.



Politics ◽  
2019 ◽  
Vol 40 (1) ◽  
pp. 3-21 ◽  
Author(s):  
Steven M Van Hauwaert ◽  
Christian H Schimpf ◽  
Flavio Azevedo

Recent research in the populism literature has devoted considerable efforts to the conceptualisation and examination of populism on the individual level, that is, populist attitudes. Despite rapid progress in the field, questions of adequate measurement and empirical evaluation of measures of populist attitudes remain scarce. Seeking to remedy these shortcomings, we apply a cross-national measurement model, using item response theory, to six established and two new populist indicators. Drawing on a cross-national survey (nine European countries, n = 18,368), we engage in a four-folded analysis. First, we examine the commonly used 6-item populism scale. Second, we expand the measurement with two novel items. Third, we use the improved 8-item populism scale to further refine equally comprehensive but more concise and parsimonious populist measurements. Finally, we externally validate these sub-scales and find that some of the proposed sub-scales outperform the initial 6- and 8-item scales. We conclude that existing measures of populism capture moderate populist attitudes, but face difficulties measuring more extreme levels, while the individual information of some of the populist items remains limited. Altogether, this provides several interesting routes for future research, both within and between countries.



2018 ◽  
Vol 28 (3) ◽  
pp. 399-429
Author(s):  
MARI C. JONES

ABSTRACTThis study examines contact-induced change in Jèrriais, the severely endangered Norman variety currently spoken by some 1% of the population of Jersey, one of the British Channel Islands. Today, English dominates all linguistic domains of island life, and all speakers of Jèrriais are bilingual. The analysis uses original data to test empirically whether Myers-Scotton's (2002) five theoretical assumptions about the structural path of language attrition (broadly defined as language loss at the level of the individual) also have relevance for the process of language obsolescence (broadly defined as language loss at the level of the community). It explores i) whether Jèrriais is undergoing contact influenced language change owing to its abstract grammatical structure being split and recombined with English, a hypothesis related to Myers-Scotton's Abstract Level model; and ii) whether different morpheme types of Jèrriais are related to the production process in different ways and are, accordingly, more or less susceptible to change during the process of language obsolescence, a hypothesis related to Myers-Scotton's 4-M model. In addition to its contribution to linguistic theory, this study increases existing knowledge about Jèrriais and makes data from this language available for systematic comparison with other languages.



Sign in / Sign up

Export Citation Format

Share Document