Deep Convolutional Neural Network for Rice Density Prescription Map at Ripening Stage Using Unmanned Aerial Vehicle-Based Remotely Sensed Images

In this paper, UAV (unmanned aerial vehicle, DJI Phantom4RTK) and YOLOv4 (You Only Look Once) target detection deep neural network methods were employed to collected mature rice images and detect rice ears to produce a rice density prescription map. The YOLOv4 model was used for rice ear quick detection of rice images captured by a UAV. The Kriging interpolation algorithm was used in ArcGIS to make rice density prescription maps. Mature rice images collected by a UAV were marked manually and used to build the training and testing datasets. The resolution of the images was 300 × 300 pixels. The batch size was 2, and the initial learning rate was 0.01, and the mean average precision (mAP) of the best trained model was 98.84%. Exceptionally, the network ability to detect rice in different health states was also studied with a mAP of 95.42% in the no infection rice images set, 98.84% in the mild infection rice images set, 94.35% in the moderate infection rice images set, and 93.36% in the severe infection rice images set. According to the severity of rice sheath blight, which can cause rice leaves to wither and turn yellow, the blighted grain percentage increased and the thousand-grain weight decreased, the rice images were divided into these four infection levels. The ability of the network model (R2 = 0.844) was compared with traditional image processing segmentation methods (R2 = 0.396) based on color and morphology features and machine learning image segmentation method (Support Vector Machine, SVM R2 = 0.0817, and K-means R2 = 0.1949) for rice ear counting. The results highlight that the CNN has excellent robustness, and can generate a wide range of rice density prescription maps.

You won’t guess that: On the limited benefits of guessing when learning a foreign language

Guessing the meaning of a foreign word before being presented with the right answer benefits recognition performance for the translation compared to reading the full translation outright. However, guessing does not increase memory for the foreign-word-to-translation associations, which is crucial for language acquisition. In this study, we aimed to investigate whether this disadvantage of guessing for performance in cued-recall tests would be eliminated if a restudy phase was added. In Experiments 1–3, we consistently demonstrated that guessing resulted in lower cued-recall performance compared to reading, both before and after restudy. Even for items for which participants successfully recalled their initial guesses on the cued-recall test, accuracy levels did not exceed those from the reading condition. In Experiment 4, we aimed to generalize our findings concerning restudy to a different set of materials – weakly associated word pairs. Even though this time guessing led to better performance than reading, consistent with previous studies, this guessing benefit was not moderated by adding a restudy phase. Our results thus underscore the importance of the initial learning phase for future learning and retention, while undermining the usefulness of the learning-through-guessing strategy for acquiring foreign language vocabulary.

Deep Learning to Detect and Classify the Purity Level of Luwak Coffee Green Beans

Luwak coffee (palm civet coffee) is known as one of the most expensive coffee in the world. In order to lower production costs, Indonesian producers and retailers often mix high-priced Luwak coffee with regular coffee green beans. However, the absence of tools and methods to classify Luwak coffee counterfeiting makes the sensing method’s development urgent. The research aimed to detect and classify Luwak coffee green beans purity into the following purity categories, very low (0-25%), low (25-50%), medium (50-75%), and high (75-100%). The classifying method relied on a low-cost commercial visible light camera and the deep learning model method. Then, the research also compared the performance of four pre-trained convolutional neural network (CNN) models consisting of SqueezeNet, GoogLeNet, ResNet-50, and AlexNet. At the same time, the sensitivity analysis was performed by setting the CNN parameters such as optimization technique (SGDm, Adam, RMSProp) and the initial learning rate (0.00005 and 0.0001). The training and validation result obtained the GoogLeNet as the best CNN model with optimizer type Adam and learning rate 0.0001, which resulted in 89.65% accuracy. Furthermore, the testing process using confusion matrix from different sample data obtained the best CNN model using ResNet-50 with optimizer type RMSProp and learning rate 0.0001, providing an accuracy average of up to 85.00%. Later, the CNN model can be used to establish a real-time, non-destructive, rapid, and precise purity detection system.

The initial learning curve for the ROSA® Knee System can be achieved in 6-11 cases for operative time and has similar 90-day complication rates with improved implant alignment compared to manual instrumentation in total knee arthroplasty

Purpose The purpose of this study was to determine the learning curve for total operative time using a novel cutting guide positioning robotic assistant for total knee arthroplasty (raTKA). Additionally, we compared complications and final limb alignment between raTKA and manual TKA (mTKA), as well as accuracy to plan for raTKA cases. Methods We performed a retrospective cohort study on a series of patients (n = 180) that underwent raTKA (n = 90) using the ROSA Total Knee System or mTKA (n = 90) by one of three high-volume (> 200 cases per year) orthopaedic surgeons between December 2019 and September 2020, with minimum three-month follow-up. To evaluate the learning curve surgical times and postoperative complications were reviewed. Results The cumulative summation analysis for total operative time revealed a change point of 10, 6, and 11 cases for each of three surgeons, suggesting a rapid learning curve. There was a significant difference in total operative times between the learning raTKA and both the mastered raTKA and mTKA groups (p = 0.001) for all three surgeons combined. Postoperative complications were minimal in all groups. The proportion of outliers for the final hip-knee-ankle angle compared to planned was 5.2% (3/58) for the mastered raTKA compared to 24.1% (19/79) for mTKA (p = 0.003). The absolute mean difference between the validated and planned resections for all angles evaluated was < 1 degree for the mastered raTKA cases. Conclusion As the digital age of medicine continues to develop, advanced technologies may disrupt the industry, but should not disrupt the care provided. This cutting guide positioning robotic system can be integrated relatively quickly with a rapid initial learning curve (6-11 cases) for operative times, similar 90-day complication rates, and improved component positioning compared to mTKA. Proficiency of the system requires additional analysis, but it can be expected to improve over time. Level of evidence Level III Retrospective Therapeutic Cohort Study.

Cortico-striatal activity characterizes human safety learning via Pavlovian conditioned inhibition

Safety learning generates associative links between neutral stimuli and the absence of threat, promoting the inhibition of fear and security-seeking behaviours. Precisely how safety learning is mediated at the level of underlying brain systems, particularly in humans, remains unclear. Here, we integrated a novel Pavlovian conditioned inhibition task with ultra-high field (UHF) fMRI to examine the neural basis of inhibitory safety learning in 49 healthy participants. In our task, participants were conditioned to two safety signals: a conditioned inhibitor that predicted threat-omission when paired with a known threat signal (A+/AX-), and a standard safety signal that generally predicted threat-omission (BC-). Both safety signals evoked equivalent autonomic and subjective learning responses but diverged strongly in terms of underlying brain activation. The conditioned inhibitor was characterized by more prominent activation of the dorsal striatum, anterior insular and dorsolateral prefrontal cortex compared to the standard safety signal, whereas the latter evoked greater activation of the ventromedial prefrontal cortex, posterior cingulate and hippocampus, among other regions. Further analyses of the conditioned inhibitor indicated that its initial learning was characterized by consistent engagement of dorsal striatal, midbrain, thalamic, premotor, and prefrontal subregions. These findings suggest that safety learning via conditioned inhibition involves a distributed cortico-striatal circuitry, separable from broader cortical regions involved with processing standard safety signals (e.g., CS-). This cortico-striatal system could represent a novel neural substrate of safety learning, underlying the initial generation of stimulus-safety associations, distinct from wider cortical correlates of safety processing, which facilitate the behavioral outcomes of learning.

A Zeroth-Order Adaptive Learning Rate Method to Reduce Cost of Hyperparameter Tuning for Deep Learning

Due to powerful data representation ability, deep learning has dramatically improved the state-of-the-art in many practical applications. However, the utility highly depends on fine-tuning of hyper-parameters, including learning rate, batch size, and network initialization. Although many first-order adaptive methods (e.g., Adam, Adagrad) have been proposed to adjust learning rate based on gradients, they are susceptible to the initial learning rate and network architecture. Therefore, the main challenge of using deep learning in practice is how to reduce the cost of tuning hyper-parameters. To address this, we propose a heuristic zeroth-order learning rate method, Adacomp, which adaptively adjusts the learning rate based only on values of the loss function. The main idea is that Adacomp penalizes large learning rates to ensure the convergence and compensates small learning rates to accelerate the training process. Therefore, Adacomp is robust to the initial learning rate. Extensive experiments, including comparison to six typically adaptive methods (Momentum, Adagrad, RMSprop, Adadelta, Adam, and Adamax) on several benchmark datasets for image classification tasks (MNIST, KMNIST, Fashion-MNIST, CIFAR-10, and CIFAR-100), were conducted. Experimental results show that Adacomp is not only robust to the initial learning rate but also to the network architecture, network initialization, and batch size.

Task compliance predicts suppression-induced forgetting in a large sample

Suppression-induced forgetting (SIF) refers to a memory impairment resulting from repeated attempts to stop the retrieval of unwanted memory associates. SIF has become established in the literature through a growing number of reports built upon the Think/No-Think (TNT) paradigm. Not all individuals and not all reported experiments yield reliable forgetting, however. Given the reliance on task instructions to motivate participants to suppress target memories, such inconsistencies in SIF may reasonably owe to differences in compliance or expectations as to whether they will again need to retrieve those items (on, say, a final test). We tested these possibilities on a large (N = 497) sample of TNT participants. In addition to successfully replicating SIF, we found that the magnitude of the effect was significantly and negatively correlated with participants’ reported compliance during the No-Think trials. This pattern held true on both same- and independent-probe measures of forgetting, as well as when the analysis was conditionalized on initial learning. In contrast, test expectancy was not associated with SIF. Supporting previous intuition and more limited post-hoc examinations, this study provides robust evidence that a lack of compliance with No-Think instructions significantly compromises SIF. As such, it suggests that diminished effects in some studies may owe, at least in part, to non-compliance—a factor that should be carefully tracked and/or controlled. Motivated forgetting is possible, provided that one is sufficiently motivated and capable of following the task instructions.

Learning from the physical consequences of our actions improves motor memory

Actions have consequences. Motor learning involves correcting actions that lead to movement errors and remembering these actions for future behavior. In most laboratory situations, movement errors have no physical consequences and simply indicate the progress of learning. Here we asked how experiencing a physical consequence when making a movement error affects motor learning. Two groups of participants adapted to a new, prism-induced mapping between visual input and motor output while performing a precision walking task. Importantly, one group experienced an unexpected slip perturbation when making foot-placement errors during adaptation. Because of our innate drive for safety, and the fact that balance is fundamental to movement, we hypothesized that this experience would enhance motor memory. Learning generalized to different walking tasks to a greater extent in the group who experienced the adverse physical consequence. This group also showed faster relearning one week later despite exposure to a competing mapping during initial learning—evidence of greater memory consolidation. The group differences in generalization and consolidation occurred even though they both experienced similar magnitude foot-placement errors and adapted at similar rates. Our results suggest the brain considers the potential physical consequences of movement error when learning and that balance-threatening consequences serve to enhance this process.

IMPLEMENTATION PLAN ON LEARNING ORGANIZATION PRACTICE TO INCREASE BUSINESS PERFORMANCE OF SAVLEE COMPANY

Cessa previously ran a technology-based business by creating therapeutic tools to reduce sleep problems. Since the pandemic era, the research and development process has been hampered due to limited access to laboratories. Therefore, the company pivoted to the Savlee company by creating a dailywear sleeping product. There are several problems and challenges that arise in the business model. Savlee began to try a new approach in the fields of pharmacy, psychology,, and neuroscience. Savlee needed to have initial learning and should seek a new form of data and information that became the main reference. The data analysis process was carried out using qualitative data methods with in-depth interviews and focus group discussions. Based on the House of Learning Organization by Jann Model as the model of learning organization practices that consist of learning foundations, learning facilities, learning skills, learning enablers, and learning disciplines, and plan for the implementing practice of learning organizations using Peter Senge's methods in "The Dance of Change-Generating Profound Change." The outcome of this research is a plan for implementation of the learning organization, which consists of three stages, starting with building individual personal mastery, team learning ability, and organizational learning ability. Therefore, Savlee must implement a suitable learning organization with appropriate design and organizational learning that has a clear measurement for the evaluation of business performance