Conditional Discrimination Research Articles

Hyperspectral Remote Sensing Imagery Generation From RGB Images Based on Joint Discrimination

Spatial resolution and spectral resolution both play an important role in the recognition of objects in hyperspectral remote sensing. However, the imaging characteristics of hyperspectral images (HSIs) result in a mutually restrictive relationship between the spatial and spectral resolutions. Generative adversarial networks (GANs) have achieved significant success in image generation. The introduce of the discriminators plays a key role in improving the reality. In this article, we propose an RGB to multiband hyperspectral imagery (150 bands) generation method based on GAN (R2HGAN). The method solves the high ill-posed problem and introduces high spectral resolution into RGB images by learning from multiple scenes of HSI. In R2HGAN, we extend the adversarial learning from spatial to spectral dimensions and joint discrimination is designed to generate HSIs closer to the real ones, where two discriminators (the conditional D and the spectral D) are put forward to supervise the spectral similarity and the conditional reality of the HSI jointly. In detail, the conditional discriminator comprehensively judges the quality of each area in the reconstructed HSI. At the same time, to ensure that the generated spectra are close to the real ones, a spectral discriminator based on multilayer perceptron is designed. Through the experiments on GF-5 imagery, the method has significantly improved the quality of the generated images over other state-of-the-art methods.

Generation of echocardiographic 2D images of the heart using cGAN

One of the most significant tasks of echocardiography is the automatic delineation of the cardiac structures from 2D echocardiographic images. Over the past decades, the automation of this task has been the subject of intense research. One of the most effective approaches is based on the deep convolutional neural networks. Nonetheless, it is necessary to use echocardiogram frames of the cardiac muscle, which show the boundaries of the cardiac structures labeled/annotated by experts/cardiologists to train it. However, the number of databases containing the necessary information is relatively small. Therefore, generated echocardiogram frames are used to increase the amount of training samples. This process is based on the ultrasound images of the heart, annotated by experts. The article proposes an improved method for generating echocardiograms using a generative adversarial neural network (GAN) with a patch-based conditional discriminator. It has been demonstrated that it is possible to improve the quality of generated echocardiogram frames in both two and four chamber views (AP4C, AP2C) using the masks of cardiac segmentation with sub-pixel convolution layer (pixel shuffle). It is demonstrated that the proposed approach makes it possible to generate ultrasound images, the structure of which corresponds to the specified segmentation masks. It is expected that this method will improve the accuracy of solving the direct problem of automatic segmentation of the left ventricle.

Age-Oriented Face Synthesis With Conditional Discriminator Pool and Adversarial Triplet Loss.

The vanilla Generative Adversarial Networks (GANs) are commonly used to generate realistic images depicting aged and rejuvenated faces. However, the performance of such vanilla GANs in the age-oriented face synthesis task is often compromised by the mode collapse issue, which may produce poorly synthesized faces with indistinguishable visual variations. In addition, recent age-oriented face synthesis methods use the L1 or L2 constraint to preserve the identity information in synthesized faces, which implicitly limits the identity permanence capabilities when these constraints are associated with a trivial weighting factor. In this paper, we propose a method for the age-oriented face synthesis task that achieves high synthesis accuracy with strong identity permanence capabilities. Specifically, to achieve high synthesis accuracy, our method tackles the mode collapse issue with a novel Conditional Discriminator Pool, which consists of multiple discriminators, each targeting one particular age category. To achieve strong identity permanence capabilities, our method uses a novel Adversarial Triplet loss. This loss, which is based on the Triplet loss, adds a ranking operation to further pull the positive embedding towards the anchor embedding to significantly reduce intra-class variances in the feature space. Through extensive experiments, we show that our proposed method outperforms state-of-the-art methods in terms of synthesis accuracy and identity permanence capabilities, both qualitatively and quantitatively.

Expansion of arbitrary stimulus classes and function-transfer measured by sorting performances.

Sorting (SRT) and matching-to-sample (MTS) tests have measured the formation of arbitrary stimulus classes. This experiment used SRT and MTS tests to document the expansion of class size. Thirty-two participants learned 12 conditional discriminations with a linear series training structure (A➔B➔C➔D➔E). SRT tests documented the formation of 5-member classes by 17 of the participants. Thereafter, 6-member class expansion was implemented by FC training. Nine of these 17 participants showed class expansion when tracked with a sequence of an SRT, MTS, and a final SRT test, and the other 8 showed expansion when tracked with a sequence of MTS and SRT tests. Thus, SRT tests documented class expansion, and the sequence of tests did not influence class expansion. The 15 participants who did not form the 5-member classes learned the baselines for new 3-member classes (A➔B➔C) and formed them as documented by an SRT test. Then, 4-member class expansion was implemented by FB training. Expansion was assessed using the above-mentioned testing sequences. All 15 showed class expansion with 100% correspondence between the SRT and MTS performances. Sorting documented the expansion of arbitrary stimulus classes, while the MTS tests showed that the stimuli also functioned as members of equivalence classes.

The Rock Burst Hazard Evaluation Using Statistical Learning Approaches

The great threat and destructiveness brought by a rock burst make its prediction and prevention crucial in engineering. The rock burst hazard evaluation at project locations is an effective way of preventing rock burst since currently real‐time prediction is not available. Since different control factors and discrimination conditions of rock burst were accepted by conventional risk determination methods, the rock burst risk determination in the same area may produce conflicting results. In this study, Naive Bayes statistical learning models based on different model prior distributions representing highly complicated nonlinear relationship between rock burst hazard and impact factors were built to evaluate the rock burst hazards. The results suggested that the Bayes statistical learning model based on a Gaussian prior has the strongest performance over four preset prior distributions. Combining the rock mechanics parameters measured in the laboratory and the stress data collected on the project sites, the proposed model was successfully employed to evaluate the kimberlite rock burst risk of a diamond mine in Canada. The Bayes statistical learning model exhibits its robustness and generalization in rock burst hazard evaluation, which can be generalized for similar engineering cases with enough supported data.

A new anisotropic continuum traffic flow model with anticipation driving behavior

Based on the anticipation driving car-following model, a new macro traffic flow model is established in this paper by considering the relationship between micro and macro variables. Therefore, the evolution law of traffic flow with anticipation driving effect can be studied from macroscopic level. By using approaches of linear stability analysis, the linear stability discriminant condition of the new macro model to keep the traffic flow stable against small disturbance is obtained. Numerical experiments verify that the model can not only simulate the unique shock wave, rarefaction wave mutation and the dynamic propagation process of small disturbance, but also improve the stability of traffic flow by introducing the information of anticipation driving behavior.

여군 인권의 실태조사 분석 연구-여군의 인권전반과 차별의 인식 및 실태를 중심으로-

여군 인권의 실태조사 분석 연구-여군의 인권전반과 차별의 인식 및 실태를 중심으로-

Neutrality of third-degree price discrimination

This note shows that compared to uniform pricing, third-degree price discrimination can be neutral. For linear demands, having all price intercepts of submarkets’ inverse demands to be the same is the necessary and sufficient condition for third-degree price discrimination to be neutral. For general demands, when all submarkets’ demands are proportional to each other, not only will all the submarkets be served under uniform pricing, but all prices before and after price discrimination will be the same.

Linear Demands and Third-Degree Price Discrimination with Some Caveats

This note shows that compared to uniform pricing, third-degree price discrimination can be neutral. For linear demands, having all price intercepts of submarkets’ inverse demands to be the same is the necessary and sufficient condition for third-degree price discrimination to be neutral. For general demands, when all submarkets’ demands are proportional to each other, not only will all the submarkets be served under uniform pricing, but all prices before and after price discrimination will be the same.

Dorsomedial striatal contributions to different forms of risk/reward decision making

Optimal decision making involving reward uncertainty is integral to adaptive goal-directed behavior. In some instances, these decisions are guided by internal representations of reward history, whereas in other situations, external cues inform a decision maker about how likely certain actions are to yield reward. Different regions of the frontal lobe form distributed networks with striatal and amygdalar regions that facilitate different types of risk/reward decision making. The dorsal medial striatum (DMS) is one key output region of the prefrontal cortex, yet there have been few preclinical studies investigating the involvement of the DMS in different forms of risk/reward decision making. The present study addressed this issue, wherein separate groups of male rats were trained on one of two tasks where they chose between a small/certain or a large/risky reward. In a probabilistic discounting task, reward probabilities changed systematically over blocks of trials (100–6.25% or 6.25–100%), requiring rats to use internal representations of reward history to guide choice. Cue-guided decision-making was assessed with a “Blackjack” task, where different auditory cues indicated the odds associated with the large/risky option (50 or 12.5%). Inactivation of the DMS with GABA agonists impaired adjustments in choice biases during probabilistic discounting, resulting in either increases or decreases in risky choice as the probabilities associated with the large/risky reward decreased or increased over a session. In comparison, DMS inactivation increased risky choices on poor-odds trials on the Blackjack task, which was associated with a reduced impact that non-rewarded choices had on subsequent choices. DMS inactivation also impaired performance of an auditory conditional discrimination. These findings highlight a previously uncharacterized role for the DMS in facilitating flexible action selection during multiple forms of risk/reward decision making.

Matching-to-sample performance in older and younger adults

ABSTRACT The present experiment studied the influence of simultaneous matching-to-sample (SMTS) or delayed matching-to-sample (DMTS 0 s) using the many-to-one (MTO) or one-to-many (OTM) training structures on the formation of equivalence classes in older and younger adults. Sixty older adults and 16 younger adults were divided into four groups: (1) SMTS-MTO, (2) SMTS-OTM, (3) DMTS 0 s-MTO, and (4) DMTS 0 s-OTM. All of the younger adults established conditional discriminations, whereas only 35 of the 60 older adults did the same. The SMTS-OTM had the lowest number of trials to mastery criterion for the older adults, and the SMTS-MTO group had the highest number of participants forming equivalence classes. In the younger adults, the differences in trials to mastery criterion were minimal, and there were no differences in the equivalence tests across groups. The older adults had a higher reaction time to comparison stimuli than the younger adults.

Efeito do pareamento ao modelo com atraso fixo ou ajustado na formação de classes de equivalência


 O presente estudo teve como objetivo investigar o efeito de diferentes tipos de pareamento ao modelo com atraso na formação de classes de equivalência com um delineamento intrassujeito. Cinco estudantes universitários foram expostos a duas condições experimentais. Na Condição Atraso Ajustado, o atraso entre a apresentação dos estímulos modelo e de comparação nos treinos aumentava de 1 em 1 s até atingir o atraso máximo de 12 s e na Condição Atraso Fixo o atraso era sempre de 12 s. O número de tentativas foi acoplado entre as condições experimentais. Após o treino de duas discriminações condicionais (AB e AC ou DE e DF), com estímulos visuais abstratos, foram realizados os testes de simetria, transitividade e equivalência. Os resultados demonstram que na Condição Atraso Fixo todos os participantes formaram classes de equivalência enquanto na Condição Atraso Ajustado apenas três. Esses resultados corroboram com a literatura ao demonstrarem que melhores desempenhos são obtidos após o treino de relações condicionais com pareamento com atraso fixo quando comparado aos desempenhos obtidos após o treino de relações condicionais com pareamento com atraso ajustado.

Sidman or statistics?

Murray Sidman's statements regarding variability, experimental control, and generality are interwoven with examples from the literature on conditional discrimination. Sidman's position was that statistical inferences from group studies produce no information about the behavior of individual subjects and that statistical treatment of individual subject data masks variability which may represent conditions that are not controlled. Sidman's work on conditional discrimination provides excellent examples of how complex discriminations should be examined in detail with accuracy levels obtained for each type of discrimination within an experiment. Sidman made important contributions to the foundation of behavior analysis with extensive basic research as well as applications of methods and principles to clinical and educational settings.

Bearing fruit: Piloting a novel judgment bias task in an American black bear.

Judgment bias tasks can reveal changes in affect in animals as a function of environmental manipulations such as provision of enrichment. We assessed affect in an American black bear across seasonal changes in availability of a mulberry bush. We used a novel judgment bias task in which the background color of a touchscreen signaled whether the left or right positioned stimulus was correct. The bear learned the conditional rule in which the correct action for the white background (choose left) resulted in three pieces of food and the correct action for the black background (choose right) resulted in one piece of food. On probe trials involving intermediate gray backgrounds, left side responses indicated optimism and right side responses indicated pessimism. Tests took place at the beginning, middle, and end of mulberry season and again nearing the end of the summer and early fall before hibernation. The bear showed the most optimistic responses during the phase involving increased opportunities for foraging on mulberry. A follow-up experiment confirmed that the bear preferred three food items over one food item, suggesting the quantity-based discrimination was in fact salient to this bear. This is the first evidence for conditional discrimination learning in a black bear, validating the task to assess changes in affect.

Expansion of Sidman's theory: The inclusion of prompt stimuli in equivalence classes.

Stimulus equivalence is defined as the ability to relate stimuli in novel ways after training in which not all of the stimuli had been directly linked to one another. Sidman (2000) suggested all elements of conditional discrimination training contingencies that result in equivalence potentially become class members. Research has demonstrated the inclusion of samples, comparisons, responses, and reinforcers in equivalence classes. Given the evidence that all elements of a conditional discrimination become part of the class, the purpose of this study was to determine if class-specific prompts would also enter into their relevant equivalence classes. Experiment 1 investigated the inclusion of prompts in an equivalence class using abstract stimuli with neurotypical students enrolled in higher education courses. Experiment 2 systematically replicated Experiment 1 using meaningful stimuli and individuals diagnosed with autism spectrum disorder. The results of both experiments demonstrated that class-specific prompts became part of equivalence classes with the other positive elements of the contingency. The results are discussed in terms of class expansion and the potential impact on equivalence-based instruction.

2_TD$DIFF]Total Harmonic Distortion (THD) based discrimination of normal, inrush and fault conditions in power transformer

• Total Harmonic Distortion (THD) based inrush and fault identification scheme for power transformer is proposed in this work. • Extensive PSCAD simulation is carried out with FFT analysis to estimate the THD during different operating condition of transformer. • Suitable THD limit is proposed for normal, inrush and fault condition of power transformer. • Validation of the suggested scheme has been carried out by software algorithm and displayed results shows satisfactory performance. • Hardware prototype has been developed and various inrush and fault cases are performed on the transformer to validate the proposed technique. The non linear characteristic of transformer magnetizing core leads more current/voltage signal distortion in the event of system disturbance/fault. The harmonic penetration in the current signal depends of the transformer core, characteristic of load and abnormal conditions. Different situations or abnormalities have different harmonics content and consequently different waveform signature. In this article, authors have utilized the harmonics content to identify normal, inrush condition and fault condition in power transformer. Total Harmonic Distortion (THD) of the current waveform for one cycle, is measured by performing Fast Fourier Transformation (FFT) analysis. Authors observed that the percentage THD is generally high in the case of inrush condition due to initial flux setup in magnetic core. Conversely, during fault condition, the %THD will be low (within a particular range) due to symmetric waveform. The magnitude of %THD below predefined range will be considered as normal condition and excess THD will be considered as inrush condition. On the other hand, if THD remains within a particular range then, it is considered as fault condition. In this article, authenticity of the presented scheme is checked by performing various test cases like initial inrush, internal fault, switching of considered transformer in faulty condition and CT saturation during fault condition. The hardware validation proves the competency of the proposed THD based discrimination scheme for transformer protection. The scheme presented here is simple and capable to solve complex problem of classification of inrush condition and fault condition.

Replicating stimulus-presentation orders in discrimination training.

Children with autism spectrum disorder (ASD) are taught conditional discriminations often during early intervention. Auditory-visual conditional discrimination (AVCD) training requires the presentation of multiple antecedent stimuli, and the order of stimulus presentation varies in the literature. This series of studies replicated previous literature on stimulus-presentation order in AVCD training. In Experiment 1, we compared sample-first and comparisons-first arrangements in 8 comparisons with 4 participants with ASD. For 3 participants, both presentations were efficacious. For 1 participant, the sample-first order was more likely to be efficacious. In Experiment 2, we added a sample-first-with-repetition arrangement and conducted 6 comparisons with 5 participants with ASD. Across comparisons, all 3 presentations were efficacious. Considerations for teaching AVCD to children with ASD and suggestions for further evaluation and examination of efficacy and efficiency are discussed.

Sequence generative adversarial nets with a conditional discriminator

The success of Generative Adversarial Networks (GANs) in image generation attracts researchers to design sequence GANs in text generation. However, the discriminators of those sequence GANs usually provide only one signal per sequence, which can not reflect detailed information, e.g. whether a token is appropriate in a sequence. In addition, maximum likelihood pre-training is typically used in those models, which is time-consuming and obscures the effects of adversarial training. To cope with these problems, we propose a new sequence GAN that consists of a conditional discriminator and a discriminator-augmented generator. The conditional discriminator provides a sequence with token-level signals. The generator is designed to approximate a discriminator-augmented distribution, which avoids pre-training. Experiments show that the conditional discriminator provides more informative guidance, and our model outperforms existing models according to metrics involving both sampling quality and sampling diversity.

Auditory–Visual stimuli: Effects on derived relations with compound stimuli

AbstractThis research explored the effect of teaching conditional discriminations with three procedures on the derivation of 36 stimuli relations (derived relations). The stimuli used consisted of three characteristics musical instruments, along with the corresponding picture. In the first experiment six university students were trained with simple stimuli and tested with compound auditory–visual samples; therefore, a one‐to‐many structure was used. In the second experiment, auditory stimuli were replaced by visual stimuli, for the samples used, for new students. A third experiment was implemented with an extra phase of training with compound stimuli for six new students. The structure of the experiments was: pretests (Xbcd–A; Xacd–B; Xabd–C; Xabc–D), training (A–B; A–C; A–D), and posttests (same as pretests). The difference between these conditions was the kind of stimuli used and a new phase of teaching used in condition 3: (Xbcd–A). The results indicate that training with simple stimuli on discriminations that include stimuli that are easy to discriminate from each other (words and sounds) is a sufficient condition for good posttest performance. However, when comparisons are made difficult (words only), participants show better performance on new tests if they have a learning history with compound stimuli.

Pre-adolescent stress disrupts adult, but not adolescent, safety learning

Anxiety disorders are highly prevalent across the lifespan, although diagnoses peak early in adolescence. As a method for inhibiting fear, safety signals have the potential to augment conventional treatments for anxiety. However, the ability to acquire and use safety signals during adolescence remains unclear. Moreover, the impact of stress on safety learning has received surprisingly little attention given that stress is a major factor preceding anxiety onset. In this study, mice were trained in a discriminative conditioning protocol to facilitate safety learning and were tested for fear inhibition using a conditioned safety signal. Next, independent groups of mice were exposed to chronic unpredictable stress (CUS) conditions between postnatal day 22 and 28, followed by tests for anxiety-like phenotypes or fear inhibition using a safety signal, performed either 24 h or five weeks following CUS. Pre-adolescent CUS reduced weight in adolescence and this effect endured into adulthood. CUS also increased specific anxiety-like behaviors in adolescence that were unique from the increase in anxiety observed in adulthood. Despite increased anxiety-like behaviors, adolescents were able to learn about and effectively use safety signals to inhibit fear. In contrast, adults that experienced CUS showed a subtle increase in anxiety but had impaired safety signal learning and usage. Together, these findings indicate that pre-adolescent stress has immediate and enduring effects on anxiety-like behaviors but impairs the capacity for conditioned inhibition only following incubation.