Table-Balancing Cooperative Robot Based on Deep Reinforcement Learning.

Reinforcement learning is one of the artificial intelligence methods that enable robots to judge and operate situations on their own by learning to perform tasks. Previous reinforcement learning research has mainly focused on tasks performed by individual robots; however, everyday tasks, such as balancing tables, often require cooperation between two individuals to avoid injury when moving. In this research, we propose a deep reinforcement learning-based technique for robots to perform a table-balancing task in cooperation with a human. The cooperative robot proposed in this paper recognizes human behavior to balance the table. This recognition is achieved by utilizing the robot's camera to take an image of the state of the table, then the table-balance action is performed afterward. Deep Q-network (DQN) is a deep reinforcement learning technology applied to cooperative robots. As a result of learning table balancing, on average, the cooperative robot showed a 90% optimal policy convergence rate in 20 runs of training with optimal hyperparameters applied to DQN-based techniques. In the H/W experiment, the trained DQN-based robot achieved an operation precision of 90%, thus verifying its excellent performance.

Agricultural Robot-Centered Recognition of Early-Developmental Pest Stage Based on Deep Learning: A Case Study on Fall Armyworm (Spodoptera frugiperda).

Accurately detecting early developmental stages of insect pests (larvae) from off-the-shelf stereo camera sensor data using deep learning holds several benefits for farmers, from simple robot configuration to early neutralization of this less agile but more disastrous stage. Machine vision technology has advanced from bulk spraying to precise dosage to directly rubbing on the infected crops. However, these solutions primarily focus on adult pests and post-infestation stages. This study suggested using a front-pointing red-green-blue (RGB) stereo camera mounted on a robot to identify pest larvae using deep learning. The camera feeds data into our deep-learning algorithms experimented on eight ImageNet pre-trained models. The combination of the insect classifier and the detector replicates the peripheral and foveal line-of-sight vision on our custom pest larvae dataset, respectively. This enables a trade-off between the robot's smooth operation and localization precision in the pest captured, as it first appeared in the farsighted section. Consequently, the nearsighted part utilizes our faster region-based convolutional neural network-based pest detector to localize precisely. Simulating the employed robot dynamics using CoppeliaSim and MATLAB/SIMULINK with the deep-learning toolbox demonstrated the excellent feasibility of the proposed system. Our deep-learning classifier and detector exhibited 99% and 0.84 accuracy and a mean average precision, respectively.

A computational study on the optimization of transcranial temporal interfering stimulation with high-definition electrodes using unsupervised neural networks.

Transcranial temporal interfering stimulation (tTIS) can focally stimulate deep parts of the brain related to specific functions using beats at two high frequencies that do not individually affect the human brain. However, the complexity and nonlinearity of the simulation limit it in terms of calculation time and optimization precision. We propose a method to quickly optimize the interfering current value of high-definition electrodes, which can finely stimulate the deep part of the brain, using an unsupervised neural network (USNN) for tTIS. We linked a network that generates the values of electrode currents to another network, which is constructed to compute the interference exposure, for optimization by comparing the generated stimulus with the target stimulus. Further, a computational study was conducted using 16 realistic head models. We also compared tTIS with transcranial alternating current stimulation (tACS), in terms of performance and characteristics. The proposed method generated the strongest stimulation at the target, even when targeting deep areas or performing multi-target stimulation. The high-definition tTISl was less affected than tACS by target depth, and mis-stimulation was reduced compared with the case of using two-pair inferential stimulation in deep region. The optimization of the electrode currents for the target stimulus could be performed in 3 min. Using the proposed USNN for tTIS, we demonstrated that the electrode currents of tTIS can be optimized quickly and accurately. Moreover, we confirmed the possibility of precisely stimulating the deep parts of the brain via transcranial electrical stimulation.

Estimation of Driver's Danger Level when Accessing the Center Console for Safe Driving.

This paper proposes a system for estimating the level of danger when a driver accesses the center console of a vehicle while driving. The proposed system uses a driver monitoring platform to measure the distance between the driver's hand and the center console during driving, as well as the time taken for the driver to access the center console. Three infrared sensors on the center console are used to detect the movement of the driver's hand. These sensors are installed in three locations: the air conditioner or heater (temperature control) button, wind direction control button, and wind intensity control button. A driver's danger level is estimated to be based on a linear regression analysis of the distance and time of movement between the driver's hand and the center console, as measured in the proposed scenarios. In the experimental results of the proposed scenarios, the root mean square error of driver H using distance and time of movement between the driver's hand and the center console is 0.0043, which indicates the best estimation of a driver's danger level.

Prediction Model for Health-Related Quality of Life of Elderly with Chronic Diseases using Machine Learning Techniques.

OBJECTIVES: The purposes of this study were to identify the factors that affect the health-related quality of life (HRQoL) of the elderly with chronic diseases and to subsequently develop from such factors a prediction model to help identify HRQoL risk groups that require intervention. METHODS: We analyzed a set of secondary data regarding 716 individuals extracted from the Korea National Health and Nutrition Examination Survey from 2008 to 2010. The statistical package of SPSS and MATLAB were used for data analysis and development of the prediction model. The algorithms used in the study were the following: stepwise logistic regression (SLR) analysis and machine learning (ML) techniques, such as decision tree, random forest, and support vector machine methods. RESULTS: FIVE FACTORS WITH STATISTICAL SIGNIFICANCE WERE IDENTIFIED FOR HRQOL IN THE ELDERLY WITH CHRONIC DISEASES: 'monthly income', 'diagnosis of chronic disease', 'depression', 'discomfort', and 'perceived health status.' The SLR analysis showed the best performance with accuracy = 0.93 and F-score = 0.49. The results of this study provide essential materials that will help formulate personalized health management strategies and develop interventions programs towards the improvement of the HRQoL for elderly people with chronic diseases. CONCLUSIONS: Our study is, to our best knowledge, the first attempt to identify the influencing factors and to apply prediction models for the HRQoL of the elderly with chronic diseases by using ML techniques as an alternative and complement to the traditional statistical approaches.

Predictors of medication adherence in elderly patients with chronic diseases using support vector machine models.

OBJECTIVES: The aim of this study was to establish a prediction model of medication adherence in elderly patients with chronic diseases and to identify variables showing the highest classification accuracy of medication adherence in elderly patients with chronic diseases using support vector machine (SVM) and conventional statistical methods, such as logistic regression (LR). METHODS: We included 293 chronic disease patients older than 65 years treated at one tertiary hospital. For the medication adherence, Morisky's self-report was used. Data were collected through face-to-face interviews. The mean age of the patients was 73.8 years. The classification process was performed with LR (SPSS ver. 20.0) and SVM (MATLAB ver. 7.12) method. RESULTS: Taking into account 16 variables as predictors, the result of applying LR and SVM classification accuracy was 71.1% and 97.3%, respectively. We listed the top nine variables selected by SVM, and the accuracy using a single variable, self-efficacy, was 72.4%. The results suggest that self-efficacy is a key factor to medication adherence among a Korean elderly population both in LR and SVM. CONCLUSIONS: Medication non-adherence was strongly associated with self-efficacy. Also, modifiable factors such as depression, health literacy, and medication knowledge associated with medication non-adherence were identified. Since SVM builds an optimal classifier to minimize empirical classification errors in discriminating between patient samples, it could achieve a higher accuracy with the smaller number of variables than the number of variables used in LR. Further applications of our approach in areas of complex diseases, treatment will provide uncharted potentials to researchers in the domains.

SICAGO: Semi-supervised cluster analysis using semantic distance between gene pairs in Gene Ontology.

SUMMARY: Despite the importance of using the semantic distance to improve the performance of conventional expression-based clustering, there are few freely available software that provides a clustering algorithm using the ontology-based semantic distances as prior knowledge. Here, we present the SICAGO (SemI-supervised Cluster Analysis using semantic distance between gene pairs in Gene Ontology) system that helps to discover the groups of genes more effectively using prior knowledge extracted from Gene Ontology. AVAILABILITY: http://ai.cau.ac.kr/sicago.html CONTACT: dwkim@cau.ac.kr

Prognosis prediction for Class III malocclusion treatment by feature wrapping method.

OBJECTIVE: To use the feature wrapping (FW) method to identify which cephalometric markers show the highest classification accuracy in prognosis prediction for Class III malocclusion and to compare the prediction accuracy between the FW method and conventional statistical methods such as discriminant analysis (DA). MATERIALS AND METHODS: The sample set consisted of 38 patients (15 boys and 23 girls, mean age 8.53 +/- 1.36 years) who were diagnosed with Class III malocclusion and received both first-phase (orthopedic) and second-phase (fixed orthodontic) treatments. Lateral cephalograms were taken before (T0) and after first-phase treatment (T1) and after second-phase treatment and retention (T2). Based on the measurements taken at the T2 stage, the patients were allocated into good (n = 20) or poor (n = 18) prognosis groups. Forty-six cephalometric variables on T0 lateral cephalograms were analyzed by the FW method to identify key determinants for discriminating between the two groups. Sequential forward search (SFS) algorism and support vector machine (SVM) were used in conjunction with the FW method to improve classification accuracy. To compare the prediction accuracy of the FW method with conventional statistical methods, DA was performed for the same data set. RESULTS: AB to mandibular plane angle ( degrees ) and A to N-perpendicular (mm) were selected as the most accurate cephalometric predictors by both the FW and DA methods. However, classification accuracy was higher with the FW method (97.2%) compared with DA (92.1%), because the FW method with SFS and SVM has a more precise classification algorithm. CONCLUSIONS: The FW method, which uses a learning algorithm, might be an effective alternative to DA for prognosis prediction.

Two-Phase chief complaint mapping to the UMLS metathesaurus in Korean electronic medical records.

The task of automatically determining the concepts referred to in chief complaint (CC) data from electronic medical records (EMRs) is an essential component of many EMR applications aimed at biosurveillance for disease outbreaks. Previous approaches that have been used for this concept mapping have mainly relied on term-level matching, whereby the medical terms in the raw text and their synonyms are matched with concepts in a terminology database. These previous approaches, however, have shortcomings that limit their efficacy in CC concept mapping, where the concepts for CC data are often represented by associative terms rather than by synonyms. Therefore, herein we propose a concept mapping scheme based on a two-phase matching approach, especially for application to Korean CCs, which uses term-level complete matching in the first phase and concept-level matching based on concept learning in the second phase. The proposed concept-level matching suggests the method to learn all the terms (associative terms as well as synonyms) that represent the concept and predict the most probable concept for a CC based on the learned terms. Experiments on 1204 CCs extracted from 15,618 discharge summaries of Korean EMRs showed that the proposed method gave significantly improved F-measure values compared to the baseline system, with improvements of up to 73.57%.