• Journal of Animal Science and Technology
• /
• v.46 no.4
• /
• pp.509-516
• /
• 2004

This study is to investigate the amount of biased estimates for heritability and genetic correlation according to data structure on marbling scores in Korean cattle. Breeding population with 5 generations were simulated by way of selection for carcass weight, Longissimus muscle area and latent values of marbling scores and random mating. Latent variables of marbling scores were categorized into five by the thresholds of 0, I, 2, and 3 SD(DSI) or seven by the thresholds of -2, -1, 0,1I, 2, and 3 SD(DS2). Variance components and genetic pararneters(Heritabilities and Genetic correlations) were estimated by restricted maximum likelihood on multivariate linear mixed animal models and by Gibbs sampling algorithms on multivariate threshold mixed animal models in DS1 and DS2. Simulation was performed for 10 replicates and averages and empirical standard deviation were calculated. Using REML, heritabilitis of marbling score were under-estimated as 0.315 and 0.462 on DS1 and DS2, respectively, with comparison of the pararneter(0.500). Otherwise, using Gibbs sampling in the multivariate threshold animal models, these estimates did not significantly differ to the parameter. Residual correlations of marbling score to other traits were reduced with comparing the parameters when using REML algorithm with assuming linear and normal distribution. This would be due to loss of information and therefore, reduced variation on marbling score. As concluding, genetic variation of marbling would be well defined if liability concepts were adopted on marbling score and implemented threshold mixed model on genetic parameter estimation in Korean cattle.

• Journal of agriculture & life science
• /
• v.51 no.6
• /
• pp.89-105
• /
• 2017

This study utilized 32,312 records of 17 linear type and 10 conformation traits(including final scores) of Hanwoo cows in the KAIA(Korea Animal Improvement Association) ('09~'10), with 60,556 animals in the pedigree file. Traits included stature, body length, strength, body depth, angularity, shank thickness, rump angle, rump length, pin bone width, thigh thickness, udder volume, teat length, teat placement, foot angle, hock angle, rear leg back view, body balance, breed characteristic, head development, forequarter quality, back line, rump, thigh development, udder development, leg line, and final score. Genetic and residual(co) variances were estimated using bi-trait pairwise analyses with EM-REML algorithm. Herd-year-classifier, year at classification, and calving stage were considered as fixed effects with classification months as a covariate. The heritability estimates ranged from 0.03(teat placement) to 0.42(body length). Rump length had the highest positive genetic correlation with pin bone width(0.96). Moreover, stature, body length, strength, and body depth had the highest positive genetic correlations with rump length, pin bone width, and thigh thickness(0.81-0.94). Stature, body length, strength, body depth, rump length, pin bone width, and thigh thickness traits also had high positive genetic correlations.

• Korean Journal of Remote Sensing
• /
• v.39 no.2
• /
• pp.207-221
• /
• 2023

This study suggests deep neural network models for estimating air temperature with Level 1B (L1B) datasets of GEO-KOMPSAT-2A (GK-2A). The temperature at 1.5 m above the ground impact not only daily life but also weather warnings such as cold and heat waves. There are many studies to assume the air temperature from the land surface temperature (LST) retrieved from satellites because the air temperature has a strong relationship with the LST. However, an algorithm of the LST, Level 2 output of GK-2A, works only clear sky pixels. To overcome the cloud effects, we apply a deep neural network (DNN) model to assume the air temperature with L1B calibrated for radiometric and geometrics from raw satellite data and compare the model with a linear regression model between LST and air temperature. The root mean square errors (RMSE) of the air temperature for model outputs are used to evaluate the model. The number of 95 in-situ air temperature data was 2,496,634 and the ratio of datasets paired with LST and L1B show 42.1% and 98.4%. The training years are 2020 and 2021 and 2022 is used to validate. The DNN model is designed with an input layer taking 16 channels and four hidden fully connected layers to assume an air temperature. As a result of the model using 16 bands of L1B, the DNN with RMSE 2.22℃ showed great performance than the baseline model with RMSE 3.55℃ on clear sky conditions and the total RMSE including overcast samples was 3.33℃. It is suggested that the DNN is able to overcome cloud effects. However, it showed different characteristics in seasonal and hourly analysis and needed to append solar information as inputs to make a general DNN model because the summer and winter seasons showed a low coefficient of determinations with high standard deviations.

• Journal of Radiation Protection and Research
• /
• v.23 no.3
• /
• pp.139-147
• /
• 1998

Purpose: Tissue inhomogeneity such as lung affects tumor dose as well as transmission dose in new concept of on-line dosimetry which estimates tumor dose from transmission dose using the new algorithm. This study was carried out to confirm accuracy of correction by tissue density in tumor dose estimation utilizing transmission dose. Methods: Cork phantom (CP, density $0.202\;gm/cm^3$) having similar density with lung parenchyme and polystyrene phantom (PP, density $1.040\;gm/cm^3$) having similar density with soft tissue were used. Dose measurement was carried out under condition simulating human chest. On simulating AP-PA irradiation, PPs with 3 cm thickness were placed above and below CP, which had thickness of 5, 10, and 20 cm. On simulating lateral irradiation, 6 cm thickness of PP was placed between two 10 cm thickness CPs additional 3 cm thick PP was placed to both lateral sides. 4, 6, and 10 MV x-ray were used. Field size was in the range of $3{\times}3$ cm through $20{\times}20$ cm, and phantom-chamber distance (PCD) was 10 to 50 cm. Above result was compared with another sets of data with equivalent thickness of PP which was corrected by density. Result: When transmission dose of PP was compared with equivalent thickness of CP which was corrected with density, the average error was 0.18 (${\pm}0.27$) % for 4 MV, 0.10 (${\pm}0.43$) % for 6 MV, and 0.33 (${\pm}0.30$) % for 10 MV with CP having thickness of 5 cm. When CP was 10 cm thick, the error was 0.23 (${\pm}0.73$) %, 0.05 (${\pm}0.57$) %, and 0.04 (${\pm}0.40$) %, while for 20 cm, error was 0.55 (${\pm}0.36$) %, 0.34 (${\pm}0.27$) %, and 0.34 (${\pm}0.18$) % for corresponding energy. With lateral irradiation model, difference was 1.15 (${\pm}1.86$) %, 0.90 (${\pm}1.43$) %, and 0.86 (${\pm}1.01$) % for corresponding energy. Relatively large difference was found in case of PCD having value of 10 cm. Omitting PCD with 10 cm, the difference was reduced to 0.47 (${\pm}$1.17) %, 0.42 (${\pm}$0.96) %, and 0.55 (${\pm}$0.77) % for corresponding energy. Conclusion When tissue inhomogeneity such as lung is in tract of x-ray beam, tumor dose could be calculated from transmission dose after correction utilizing tissue density.

• Journal of Intelligence and Information Systems
• /
• v.25 no.1
• /
• pp.163-177
• /
• 2019

As smartphones are getting widely used, human activity recognition (HAR) tasks for recognizing personal activities of smartphone users with multimodal data have been actively studied recently. The research area is expanding from the recognition of the simple body movement of an individual user to the recognition of low-level behavior and high-level behavior. However, HAR tasks for recognizing interaction behavior with other people, such as whether the user is accompanying or communicating with someone else, have gotten less attention so far. And previous research for recognizing interaction behavior has usually depended on audio, Bluetooth, and Wi-Fi sensors, which are vulnerable to privacy issues and require much time to collect enough data. Whereas physical sensors including accelerometer, magnetic field and gyroscope sensors are less vulnerable to privacy issues and can collect a large amount of data within a short time. In this paper, a method for detecting accompanying status based on deep learning model by only using multimodal physical sensor data, such as an accelerometer, magnetic field and gyroscope, was proposed. The accompanying status was defined as a redefinition of a part of the user interaction behavior, including whether the user is accompanying with an acquaintance at a close distance and the user is actively communicating with the acquaintance. A framework based on convolutional neural networks (CNN) and long short-term memory (LSTM) recurrent networks for classifying accompanying and conversation was proposed. First, a data preprocessing method which consists of time synchronization of multimodal data from different physical sensors, data normalization and sequence data generation was introduced. We applied the nearest interpolation to synchronize the time of collected data from different sensors. Normalization was performed for each x, y, z axis value of the sensor data, and the sequence data was generated according to the sliding window method. Then, the sequence data became the input for CNN, where feature maps representing local dependencies of the original sequence are extracted. The CNN consisted of 3 convolutional layers and did not have a pooling layer to maintain the temporal information of the sequence data. Next, LSTM recurrent networks received the feature maps, learned long-term dependencies from them and extracted features. The LSTM recurrent networks consisted of two layers, each with 128 cells. Finally, the extracted features were used for classification by softmax classifier. The loss function of the model was cross entropy function and the weights of the model were randomly initialized on a normal distribution with an average of 0 and a standard deviation of 0.1. The model was trained using adaptive moment estimation (ADAM) optimization algorithm and the mini batch size was set to 128. We applied dropout to input values of the LSTM recurrent networks to prevent overfitting. The initial learning rate was set to 0.001, and it decreased exponentially by 0.99 at the end of each epoch training. An Android smartphone application was developed and released to collect data. We collected smartphone data for a total of 18 subjects. Using the data, the model classified accompanying and conversation by 98.74% and 98.83% accuracy each. Both the F1 score and accuracy of the model were higher than the F1 score and accuracy of the majority vote classifier, support vector machine, and deep recurrent neural network. In the future research, we will focus on more rigorous multimodal sensor data synchronization methods that minimize the time stamp differences. In addition, we will further study transfer learning method that enables transfer of trained models tailored to the training data to the evaluation data that follows a different distribution. It is expected that a model capable of exhibiting robust recognition performance against changes in data that is not considered in the model learning stage will be obtained.