Deep Learning based Handwriting Recognition with Adversarial Feature Deformation and Regularization

Due to the complex and irregular shapes of handwritten text, it is challenging to spot and recognize the handwritten words. In low-resource scripts, retrieval of words is a difficult and laborious task. The need for increasing the number of samples and introducing variations in the extended training datasets occur with the use of deep learning and neural network models. All possible variations and occurrences cannot be covered in an efficient manner with the use of the existing preprocessing strategies and theories. A scalable and elastic methodology for wrapping the extracted features is presented with the introduction of an adversarial feature deformation and regularization module in this paper. In the original deep learning framework, this module is introduced between the intermediate layers while training in an alternative manner. When compared to the conventional models, highly informative features are learnt in an efficient manner with the help of this setup. Extensive word datasets are used for testing the proposed model, which is built on popular frameworks available for word recognition and spotting, while enhancing them with the proposed module. While varying the training data size, the results are recorded and compared with the conventional models. Improvement in the mAP scores, word-error rate and low data regime is observed from the results of comparison.

Training-Dependent Change in Content of Association in Appetitive Pavlovian Conditioning

In appetitive Pavlovian conditioning, experience with a conditional relationship between a cue [conditioned stimulus (CS)] and a reward [unconditioned stimulus (US)] bestows CS with the ability to promote adaptive behavior patterns. Different features of US (e.g., identity-specific sensory, general motivational) can be encoded by CS based on the nature of the CS-US relationship experienced (e.g., temporal factors such as training amount) and the content of association may determine the influence of CS over behavior (e.g., mediated learning, conditioned reinforcement). The content of association changed with varying conditioning factors, thereby altering behavioral consequences, however, has never been addressed in relevant brain signals evoked by CS. Our previous study found that phospholipase C β1-knockout (PLCβ1-KO) mice display persistent mediated learning over the extended course of odor-sugar conditioning, and that wild-type (WT) mice lose mediated learning sensitivity after extended training. In this study, in order to see whether this behavioral difference between these two genotypes comes from a difference in the course of association content, we examined whether odor CS can evoke the taste sensory representation of an absent sugar US after minimal- and extended training in these mice. In contrast to WT, which lost CS-evoked neural activation (c-Fos expression) in the gustatory cortex after extended training, KO mice displayed persistent association with the sensory feature of sugar, suggesting that sensory encoding is reliably linked to mediated learning sensitivity and there is a training-dependent change in the content of association in WT. PLCβ1 knockdown in the left medial prefrontal cortex (mPFC) resulted in mediated learning sensitivity and CS-evoked gustatory cortical activation after extended training, proposing a molecular component of the neural system underlying this Pavlovian conditioning process. We also discuss how disruption of this process is implicated for hallucination-like behaviors (impaired reality testing).

Maintained goal-directed control with overtraining on ratio schedules

It is thought that goal-directed control of actions weakens or becomes masked by habits over time. We tested the opposing hypothesis that goal-directed control becomes stronger over time, and that this growth is modulated by the overall action–outcome contiguity. Despite group differences in action–outcome contiguity early in training, rats trained under random and fixed ratio schedules showed equivalent goal-directed control of lever pressing that appeared to grow over time. We confirmed that goal-directed control was maintained after extended training under another type of ratio schedule—continuous reinforcement—using specific satiety and taste aversion devaluation methods. These results add to the growing literature showing that extensive training does not reliably weaken goal-directed control and that it may strengthen it, or at least maintain it.

Investigating habits in humans with a symmetrical outcome revaluation task

To investigate the balance between goal-directed and habitual control in controlled experimental settings, animal researchers developed the outcome-revaluation paradigm. The translation of this paradigm to humans has yielded interesting insights but proven to be challenging. We present a novel, symmetrical outcome-revaluation task in which outcomes are both devalued and upvalued to reveal the disadvantage and advantage of habit formation. During the instrumental learning phase, participants learned to respond (Go) to certain stimuli to collect valuable outcomes (and points) while refraining to respond (NoGo) to stimuli signalling not-valuable outcomes. Half of the stimuli were short-trained, while the other half were long-trained. Subsequently, in the test phase, the signalled outcomes were either value-congruent with training (still-valuable and still-not-valuable), or incongruent (devalued and upvalued). The change in outcome value on incongruent trials meant that participants had to flexibly adjust their behaviour. At the end of the training phase, participants completed the self-report behavioural automaticity index – providing an automaticity score for each stimulus-response association. We conducted two experiments using this task, that both provided evidence for stimulus-driven habits as reflected in better performance on congruent than on incongruent test trials. While self-reported automaticity increased with longer training, behavioural flexibility was intact. After extended training (Experiment 2), higher levels of self-reported automaticity when responding to stimuli signalling valuable outcomes was related to more ‘slips of action’ when the associated outcome was subsequently devalued. We conclude that the symmetrical outcome revaluation task provides a promising paradigm for the experimental investigation of habits in humans.

Characterization of striatal dopamine projections across striatal subregions in behavioral flexibility

Behavioral flexibility is key to survival in a dynamic environment. While flexible, goal-directed behaviors are initially dependent on dorsomedial striatum, they become dependent on lateral striatum with extended training as behaviors become inflexible. Similarly, dopamine release shifts from ventromedial to lateral striatum across learning, and impairment of lateral dopamine release disrupts habitual, inflexible responding. This raises the possibility that lateral dopamine release is a causative mechanism in establishing inflexible behaviors late in training, though this has not been directly tested. Here, we utilized optogenetics to activate dopamine terminals in dorsal medial (DMS), dorsal lateral (DLS), and ventral (NAc) striatum in DATcre mice to determine how specific dopamine subpopulations impact behavioral flexibility. Mice performed a reversal task in which they self-stimulated DMS, DLS, or NAc dopamine terminals by pressing one of two levers before action-outcome lever contingencies were reversed. Consistent with presumed ventromedial/lateral striatal function, we found that mice self-stimulating ventromedial dopamine terminals rapidly reversed lever preference following contingency reversal, while mice self-stimulating dopamine terminals in DLS showed impaired reversal learning. These impairments were characterized by more regressive errors and reliance on lose-stay strategies following reversal, suggesting reward insensitivity and overreliance on previously learned actions. This study supports a model of striatal function in which dorsomedial dopamine facilitates goal-directed responding, and dorsolateral dopamine release is a key mechanism in supporting the transition toward inflexible behaviors.

Regulation of auditory fear discrimination by the novel Kv3 voltage-gated potassium channel modulator AUT00206

Psychiatric diseases like anxiety-related disorders and schizophrenia are characterized by impaired cognition and emotional regulation linked to corticolimbic disinhibition. Restoring the balance between excitation and inhibition in corticolimbic circuits may therefore ameliorate certain features of these disorders, such as inappropriately attributing affective salience to innocuous cues. Corticolimbic activity is tightly controlled by parvalbumin-expressing GABAergic interneurons, which also regulate fear discrimination. The voltage-gated potassium channels Kv3.1 and Kv3.2 are highly expressed in these neurons, therefore Kv3.1/3.2 modulation may have potential for treating disorders associated with cognitive and emotional dysregulation. We determined the effects of the novel Kv3.1/3.2 positive modulator AUT00206 on fear discrimination. Female rats underwent limited or extended auditory fear discrimination training that we previously showed leads to discrimination or generalization, respectively, based on passive fear responding (i.e. freezing). We also assessed darting as an active fear response. We found that limited training resulted in discrimination based on freezing, which was unaffected by AUT00206. In contrast, we found that extended training resulted in generalization based on freezing and the emergence of discrimination based on darting. Importantly, AUT00206 had dissociable effects on fear discrimination and expression with extended training. While AUT00206 mitigated generalization without affecting expression based on freezing, it reduced expression without affecting discrimination based on darting. Our results indicate that Kv3.1/3.2 modulation regulates the attribution of affective significance to threat- and safety-related cues in a response-specific manner. This suggests that targeting Kv3.1 and Kv3.2 channels may provide a promising avenue for treating cognitive and emotional dysregulation in psychiatric disease.

What Is Learned in Pavlovian Conditioning in Crickets? Revisiting the S-S and S-R Learning Theories

In Pavlovian conditioning in mammals, two theories have been proposed for associations underlying conditioned responses (CRs). One theory, called S-S theory, assumes an association between a conditioned stimulus (CS) and internal representation of an unconditioned stimulus (US), allowing the animal to adjust the CR depending on the current value of the US. The other theory, called S-R theory, assumes an association or connection between the CS center and the CR center, allowing the CS to elicit the CR. Whether these theories account for Pavlovian conditioning in invertebrates has remained unclear. In this article, results of our studies in the cricket Gryllus bimaculatus are reviewed. We showed that after a standard amount of Pavlovian training, crickets exhibited no response to odor CS when water US was devalued by providing it until satiation, whereas after extended training, they exhibited a CR after US devaluation. An increase of behavioral automaticity by extended training has not been reported in Pavlovian conditioning in any other animals, but it has been documented in instrumental conditioning in mammals. Our pharmacological analysis suggested that octopamine neurons mediate US (water) value signals and control execution of the CR after standard training. The control, however, diminishes with extension of training and hence the CR becomes insensitive to the US value. We also found that the nature of the habitual response after extended Pavlovian training in crickets is not the same as that after extended instrumental training in mammals concerning the context specificity. Adaptive significance and evolutionary implications for our findings are discussed.

Using IRB Protocols to Teach Ethical Principles for Research and Everyday Life

Undergraduate research as a high-impact practice demonstrates many positive benefits for students, but little research has delved into the impact of ethical training for research, in particular submitting Institutional Review Board (IRB) protocols to determine if the study meets ethical standards for the treatment of human subjects. This study explored if students in two experimental and one nonexperimental research methods class benefited from increased knowledge of research ethics and how to apply them in daily-life situations if they participated in various aspects of IRB protocol procedures either as part of a class-based research project or by completing an IRB protocol activity for developing a hypothetical program to help families. Some students in all three classes had previously engaged in a 4-hr online extended training [the Collaborative Institutional Training Initiative (CITI) Program] in research ethics focused on the Belmont Report principles of beneficence, respect, and justice, but not in IRB protocols. Students were given a pre- and posttest to assess knowledge in both research and daily-life settings for applying the Belmont Report research ethics principles. Results indicate students gained greater knowledge of research ethics when they completed IRB protocol training during a class-based undergraduate research or program-design project, even if they had already completed some extended case-based training in the CITI Program. Results are discussed in terms of the value of using modified IRB protocol approaches as a high-impact practice to teach ethics in research and daily life to students.

Probing the decision-making mechanisms underlying choice between drug and nondrug rewards in rats

Delineating the decision-making mechanisms underlying choice between drug and nondrug rewards remains a challenge. This study adopts an original approach to probe these mechanisms by comparing response latencies during sampling versus choice trials. While lengthening of latencies during choice is predicted in a deliberative choice model (DCM), the race-like response competition mechanism postulated by the Sequential choice model (SCM) predicts a shortening of latencies during choice compared to sampling. Here, we tested these predictions by conducting a retrospective analysis of cocaine-versus-saccharin choice experiments conducted in our laboratory. We found that rats engage deliberative decision-making mechanisms after limited training, but adopt a SCM-like response selection mechanism after more extended training, while their behavior is presumably habitual. Thus, the DCM and SCM may not be general models of choice, as initially formulated, but could be dynamically engaged to control choice behavior across early and extended training.