scholarly journals EvidentialMix: Learning with Combined Open-set and Closed-set Noisy Labels

Author(s):  
Ragav Sachdeva ◽  
Filipe R. Cordeiro ◽  
Vasileios Belagiannis ◽  
Ian Reid ◽  
Gustavo Carneiro
Keyword(s):  
Open Set ◽  
2020 ◽  
Vol 9 (11) ◽  
pp. 9353-9360
Author(s):  
G. Selvi ◽  
I. Rajasekaran

This paper deals with the concepts of semi generalized closed sets in strong generalized topological spaces such as $sg^{\star \star}_\mu$-closed set, $sg^{\star \star}_\mu$-open set, $g^{\star \star}_\mu$-closed set, $g^{\star \star}_\mu$-open set and studied some of its basic properties included with $sg^{\star \star}_\mu$-continuous maps, $sg^{\star \star}_\mu$-irresolute maps and $T_\frac{1}{2}$-space in strong generalized topological spaces.


2021 ◽  
Vol 2021 (1) ◽  
Author(s):  
Clara Borrelli ◽  
Paolo Bestagini ◽  
Fabio Antonacci ◽  
Augusto Sarti ◽  
Stefano Tubaro

AbstractSeveral methods for synthetic audio speech generation have been developed in the literature through the years. With the great technological advances brought by deep learning, many novel synthetic speech techniques achieving incredible realistic results have been recently proposed. As these methods generate convincing fake human voices, they can be used in a malicious way to negatively impact on today’s society (e.g., people impersonation, fake news spreading, opinion formation). For this reason, the ability of detecting whether a speech recording is synthetic or pristine is becoming an urgent necessity. In this work, we develop a synthetic speech detector. This takes as input an audio recording, extracts a series of hand-crafted features motivated by the speech-processing literature, and classify them in either closed-set or open-set. The proposed detector is validated on a publicly available dataset consisting of 17 synthetic speech generation algorithms ranging from old fashioned vocoders to modern deep learning solutions. Results show that the proposed method outperforms recently proposed detectors in the forensics literature.


2021 ◽  
Vol 11 (1) ◽  
Author(s):  
Adam Goodwin ◽  
Sanket Padmanabhan ◽  
Sanchit Hira ◽  
Margaret Glancey ◽  
Monet Slinowsky ◽  
...  

AbstractWith over 3500 mosquito species described, accurate species identification of the few implicated in disease transmission is critical to mosquito borne disease mitigation. Yet this task is hindered by limited global taxonomic expertise and specimen damage consistent across common capture methods. Convolutional neural networks (CNNs) are promising with limited sets of species, but image database requirements restrict practical implementation. Using an image database of 2696 specimens from 67 mosquito species, we address the practical open-set problem with a detection algorithm for novel species. Closed-set classification of 16 known species achieved 97.04 ± 0.87% accuracy independently, and 89.07 ± 5.58% when cascaded with novelty detection. Closed-set classification of 39 species produces a macro F1-score of 86.07 ± 1.81%. This demonstrates an accurate, scalable, and practical computer vision solution to identify wild-caught mosquitoes for implementation in biosurveillance and targeted vector control programs, without the need for extensive image database development for each new target region.


Author(s):  
Tetiana Osipchuk

The topological properties of classes of generally convex sets in multidimensional real Euclidean space $\mathbb{R}^n$, $n\ge 2$, known as $m$-convex and weakly $m$-convex, $1\le m<n$, are studied in the present work. A set of the space $\mathbb{R}^n$ is called \textbf{\emph{$m$-convex}} if for any point of the complement of the set to the whole space there is an $m$-dimensional plane passing through this point and not intersecting the set. An open set of the space is called \textbf{\emph{weakly $m$-convex}}, if for any point of the boundary of the set there exists an $m$-dimensional plane passing through this point and not intersecting the given set. A closed set of the space is called \textbf{\emph{weakly $m$-convex}} if it is approximated from the outside by a family of open weakly $m$-convex sets. These notions were proposed by Professor Yuri Zelinskii. It is known the topological classification of (weakly) $(n-1)$-convex sets in the space $\mathbb{R}^n$ with smooth boundary. Each such a set is convex, or consists of no more than two unbounded connected components, or is given by the Cartesian product $E^1\times \mathbb{R}^{n-1}$, where $E^1$ is a subset of $\mathbb{R}$. Any open $m$-convex set is obviously weakly $m$-convex. The opposite statement is wrong in general. It is established that there exist open sets in $\mathbb{R}^n$ that are weakly $(n-1)$-convex but not $(n-1)$-convex, and that such sets consist of not less than three connected components. The main results of the work are two theorems. The first of them establishes the fact that for compact weakly $(n-1)$-convex and not $(n-1)$-convex sets in the space $\mathbb{R}^n$, the same lower bound for the number of their connected components is true as in the case of open sets. In particular, the examples of open and closed weakly $(n-1)$-convex and not $(n-1)$-convex sets with three and more connected components are constructed for this purpose. And it is also proved that any compact weakly $m$-convex and not $m$-convex set of the space $\mathbb{R}^n$, $n\ge 2$, $1\le m<n$, can be approximated from the outside by a family of open weakly $m$-convex and not $m$-convex sets with the same number of connected components as the closed set has. The second theorem establishes the existence of weakly $m$-convex and not $m$-convex domains, $1\le m<n-1$, $n\ge 3$, in the spaces $\mathbb{R}^n$. First, examples of weakly $1$-convex and not $1$-convex domains $E^p\subset\mathbb{R}^p$ for any $p\ge3$, are constructed. Then, it is proved that the domain $E^p\times\mathbb{R}^{m-1}\subset\mathbb{R}^n$, $n\ge 3$, $1\le m<n-1$, is weakly $m$-convex and not $m$-convex.


1970 ◽  
Vol 13 (4) ◽  
pp. 839-855 ◽  
Author(s):  
Daniel L. Bode ◽  
Herbert J. Oyer

Thirty-two adults with sensorineural hearing loss participated in a short-term auditory training program. The listeners were assigned to one of four matched groups which were equivalent in pure-tone sensitivity, speech-reception threshold, PB discrimination in quiet and in noise, intelligence, age, education, duration of loss, sex, and hearing-aid use. Each group responded during training to a different combination of listening condition (S/N varied or S/N-constant) and speech material (closed-set or open-set response formats). Statistically significant increase in auditory discrimination was shown on the W-22 and Rhyme tests, while the increase revealed by the Semi-Diagnostic test was not significant. Results indicated that the two listening conditions were equally effective. Similarly, the two types of training material brought about equivalent increases in overall speech discrimination. Trends suggested that open-set and closed-set training each had most effect on the respective type of speech discrimination. In addition, improvement in auditory discrimination was associated with those individuals who were oldest, with those who had highest intelligence, and with those who responded to training material at the most intense presentation level. Finally, listeners who reported the most hearing handicap also tended to show the greatest loss in speech reception and in speech discrimination in noise.


1999 ◽  
Vol 106 (4) ◽  
pp. 2177-2177 ◽  
Author(s):  
Ted A. Meyer ◽  
Mario A. Svirsky ◽  
Stefan Frisch ◽  
Adam R. Kaiser ◽  
David B. Pisoni ◽  
...  

1991 ◽  
Vol 34 (5) ◽  
pp. 1180-1184 ◽  
Author(s):  
Larry E. Humes ◽  
Kathleen J. Nelson ◽  
David B. Pisoni

The Modified Rhyme Test (MRT), recorded using natural speech and two forms of synthetic speech, DECtalk and Votrax, was used to measure both open-set and closed-set speech-recognition performance. Performance of hearing-impaired elderly listeners was compared to two groups of young normal-hearing adults, one listening in quiet, and the other listening in a background of spectrally shaped noise designed to simulate the peripheral hearing loss of the elderly. Votrax synthetic speech yielded significant decrements in speech recognition compared to either natural or DECtalk synthetic speech for all three subject groups. There were no differences in performance between natural speech and DECtalk speech for the elderly hearing-impaired listeners or the young listeners with simulated hearing loss. The normal-hearing young adults listening in quiet out-performed both of the other groups, but there were no differences in performance between the young listeners with simulated hearing loss and the elderly hearing-impaired listeners. When the closed-set identification of synthetic speech was compared to its open-set recognition, the hearing-impaired elderly gained as much from the reduction in stimulus/response uncertainty as the two younger groups. Finally, among the elderly hearing-impaired listeners, speech-recognition performance was correlated negatively with hearing sensitivity, but scores were correlated positively among the different talker conditions. Those listeners with the greatest hearing loss had the most difficulty understanding speech and those having the most trouble understanding natural speech also had the greatest difficulty with synthetic speech.


1989 ◽  
Vol 32 (4) ◽  
pp. 887-911 ◽  
Author(s):  
Richard S. Tyler ◽  
Brian C. J. Moore ◽  
Francis K. Kuk

The main purpose of this study was to provide an independent corroboration of open-set word recognition in some of the better cochlear-implant patients. These included the Chorimac, Nucleus (one group from the U.S.A. and one group from Hannover, Germany), Symbion, Duren/Cologne and 3M/Vienna implants. Three experiments are reported: (1) word recognition in word lists and in sentences; (2) environmental sound perception, and (3) gap detection. On word recognition, the scores of 6 Chorimac patients averaged 2.5% words and 0.7% words in sentences correct in the French tests. In the German tests, the scores averaged 17% words and 10% words in sentences for 10 Duren/Cologne patients, 15% words and 16% words in sentences for 9 3M/Vienna patients, and 10% words and 16% words in sentences (3% to 26%) for 10 Nucleus/Hannover patients. In the English tests, the scores averaged 11% words and 29.6% words in sentences for l0 Nucleus-U.S.A. patients, and 13.7% words and 35.7% words in sentences for the 9 Symbion patients. The ability to recognize recorded environmental sounds was measured with a closed set of 18 sounds. Performance averaged 23% correct for Chorimac patients, 41% correct for 3M/Vienna patients, 44% correct for Nucleus/Hannover patients, 21% correct for Duren/Cologne patients, 58% correct for Nucleus/U.S.A. patients, and 83% correct for Symbion patients. A multidimensional scaling analysis suggested that patients were, in part, utilizing information about the envelope and about the periodic/aperiodic nature of some of the sounds. Gap detection thresholds with a one-octave wide noise centered at 500 Hz varied widely among patients. Typically, patients with gap thresholds less than 40 ms showed a wide range of performance on speech perception tasks, whereas patients with gap-detection thresholds greater than 40 ms showed poor word recognition skills.


2016 ◽  
Vol 7 (3) ◽  
pp. 145
Author(s):  
N. Durga Devia ◽  
Raja Rajeswari ◽  
P. Thangavelu

The aim of this paper is to study how distinct points and a point and a closed set not containing that points are separated by non overlapping open neighborhoods, in a bitopological space. The separation is studied with respect to a new type of \((1,2)\alpha\)-open set together with a continuous function. We named the new axioms as star-ultra \(T_{1}\), star-ultra \(T_{2}\), star-ultra regular and normal. The star-ultra regular spaces is studied in two different ways and are called as A-star-ultra regular and B-star-ultra regular spaces.


2013 ◽  
Vol 24 (10) ◽  
pp. 969-979 ◽  
Author(s):  
Min-Yu Cheng ◽  
Jaclyn B. Spitzer ◽  
Valeriy Shafiro ◽  
Stanley Sheft ◽  
Dean Mancuso

Purpose: The goals of this study were (1) to investigate the reliability of a clinical music perception test, Appreciation of Music in Cochlear Implantees (AMICI), and (2) examine associations between the perception of music and speech. AMICI was developed as a clinical instrument for assessing music perception in persons with cochlear implants (CIs). The test consists of four subtests: (1) music versus environmental noise discrimination, (2) musical instrument identification (closed-set), (3) musical style identification (closed-set), and (4) identification of musical pieces (open-set). To be clinically useful, it is crucial for AMICI to demonstrate high test-retest reliability, so that CI users can be assessed and retested after changes in maps or programming strategies. Research Design: Thirteen CI subjects were tested with AMICI for the initial visit and retested again 10–14 days later. Two speech perception tests (consonant-nucleus-consonant [CNC] and Bamford-Kowal-Bench Speech-in-Noise [BKB-SIN]) were also administered. Data Analysis: Test-retest reliability and equivalence of the test's three forms were analyzed using paired t-tests and correlation coefficients, respectively. Correlation analysis was also conducted between results from the music and speech perception tests. Results: Results showed no significant difference between test and retest (p > 0.05) with adequate power (0.9) as well as high correlations between the three forms (Forms A and B, r = 0.91; Forms A and C, r = 0.91; Forms B and C, r = 0.95). Correlation analysis showed high correlation between AMICI and BKB-SIN (r = −0.71), and moderate correlation between AMICI and CNC (r = 0.4). Conclusions: The study showed AMICI is highly reliable for assessing musical perception in CI users.


Sign in / Sign up

Export Citation Format

Share Document