Abstract: The present invention relates to a context-based QA generation architecture for generating diverse QA pairs from a single context. The context-based QA generation architecture includes a latent variable generating network, an answer generating network and a question generating network. The latent variable generating network comprises multiple Bi-LSTM encoders encode the a first context, the a first question and the a first answer to generate a first context vector, a first question vector and a first answer vector, respectively, a first Multi-Layer Perceptron (MLP) generate a first question latent variable based on the first context vector and the first question vector, and a second MLP generate a first answer latent variable based on the first question latent variable and the first answer vector. The answer generating network and the question generating network are trained based on the first context, the first question latent variable and the first answer latent variable.

Abstract: As an aspect of the present disclosure, a method of processing an image may be proposed. The method is a method of processing an image, which is performed by an electronic device including one or more processors and one or more memories in which instructions to be executed by the one or more processors are stored, and may include obtaining a first mixed image of a sample including a first biomolecule labeled with a first fluorescent material and a second biomolecule that has not been labeled, obtaining a second mixed image of the sample including the first biomolecule labeled with the first fluorescent material and the second biomolecule labeled with a second fluorescent material, and generating an unmixed image of the second biomolecule based on the first mixed image and the second mixed image.

Abstract: The present disclosure relates to a knot-architectured fabric actuator using a knot shape and an application using the same. The knot-architectured fabric actuator includes at least one knot including a node formed by fiber strands crossing each other, and at least one petal including a closed curve formed by fibers extending from both ends of the knot. Further, the knot-architectured fabric actuator may be used as a wearable robot, a soft actuator, an adaptive sleeve, or the like.

Abstract: Disclosed is a method and system for separating and recalibrating a feature for adversarial robustness. A feature separation and recalibration method may include disentangling an input feature map into a first feature defined as a robust feature and a second feature defined as a non-robust feature, based on a robustness score; generating a recalibrated feature by recalibrating activation of the second feature; and generating an output feature map by combining the first feature and the recalibrated feature.

Abstract: A method comprises obtaining original data representing one or more characteristics of a target object; selecting an artificial intelligence generation model from among a plurality of artificial intelligence generation models based on the original data; generating reconstructed data based on the original data using the selected artificial intelligence generation model; calculating an anomaly score by comparing the original data and the reconstructed data; and outputting detection result data indicating whether the target object is abnormal based on the anomaly score.

Abstract: Disclosed are an unsupervised learning method and an apparatus therefor applicable to general inverse problems. An unsupervised learning method applicable to inverse problems includes receiving a training data set and training an unsupervised learning-based neural network generated based on an optimal transport theory and a penalized least square (PLS) approach using the training data set, wherein the receiving of the training data set includes receiving the training data set including unmatched data.

Abstract: A method of operating a solid oxide fuel cell system may comprise contacting a cathode gas comprising oxygen with a heating element to produce a heated cathode gas, passing the heated cathode gas through a cathode of a solid oxide fuel cell stack to increase the temperature of the solid oxide fuel cell stack to an operation temperature and reduce the oxygen to oxygen anions, and passing an anode gas through an anode of the solid oxide fuel cell stack to initiate the electrochemical oxidation of the oxygen anions within the anode. The passing of the anode gas through the anode of the solid oxide fuel cell stack may be initiated when the solid oxide fuel cell stack is heated to an operational temperature.

Abstract: Provided herein are a method for modulation of a variable binary offset carrier (VBOC) having a discrete time-variant frequency and a satellite navigation signal generator using the same. The method for modulation of a VBOC having a discrete time-variant frequency may comprise: generating navigation message data; generating a carrier; generating a VBOC subcarrier; generating a pseudo-random code; synthesizing the pseudo-random code and the VBOC subcarrier; and synthesizing a synthesized signal of the pseudo-random code and the VBOC subcarrier, the carrier, and the navigation message data.

Abstract: Disclosed decoding method of the intra prediction mode comprises the steps of: determining whether an intra prediction mode of a present prediction unit is the same as a first candidate intra prediction mode or as a second candidate intra prediction mode on the basis of 1-bit information; and determining, among said first candidate intra prediction mode and said second candidate intra prediction mode, which candidate intra prediction mode is the same as the intra prediction mode of said present prediction unit on the basis of additional 1-bit information, if the intra prediction mode of the present prediction unit is the same as at least either the first candidate intra prediction mode or the second candidate intra prediction mode, and decoding the intra prediction mode of the present prediction unit.

Abstract: A method for protecting the privacy of personal location information by level of sensitivity and efficiently recovering a source location is disclosed. The sensitive information protection and recovery method performed by means of a privacy control system, according to one embodiment, may comprise the steps of: processing privacy protection for sensitive information of a user, adjusted on the basis of privacy level selected according to a service by which the sensitive information was requested; and recovering the privacy-protected sensitive information of the user by using a secret key generated on the basis of the sensitive information of the user, adjusted on the basis of the privacy level.

Abstract: Disclosed are a growth structure for a strained channel, and a strained channel using the same and a method of manufacturing a device using the same. The growth structure for a strained channel includes a support substrate, a strain-relaxed buffer (SRB) layer disposed on a support substrate, a base growth layer grown to have one composition on the SRB layer, and a strained channel layer grown to have another composition on the base growth layer. The strained channel layer may include at least one of a tensile-strained channel layer or a compressively strained channel layer.

Abstract: A method of accelerating training of a low-power, high-performance artificial neural network (ANN) includes (a) performing fine-grained pruning and coarse-grained pruning to generate sparsity in weights by a pruning unit in a convolution core of a cluster in a lower-power, high-performance ANN trainer; (b) selecting and performing dual zero skipping according to input sparsity, output sparsity, and the sparsity of weights by the convolution core, and (c) restricting access to a weight memory during training by allowing a deep neural network (DNN) computation core and a weight pruning core to share weights retrieved from a memory by the convolution core.

Abstract: A system and method for perceiving a 3-D environment by using a camera and a radar sensor are disclosed. In order to generate a 3-D feature map that is used to perceive an environment through the fusion of a camera and a radar sensor, the method of perceiving a 3-D environment may include extracting a two-dimensional (2-D) feature map from an image obtained by the camera and transforming the 2-D feature map into a feature map in a 3-D space by using first distance information extracted from the image and second distance information measured by the radar sensor.

Abstract: A surface-enhanced Raman spectroscopy substrate includes a multi-stage nanolattice structure including a first nanolattice including silver (Ag), a second nanolattice including gold (Au), and a third nanolattice including platinum (Pt). The surface-enhanced Raman spectroscopy substrate may be used for detecting different target materials.

Abstract: The present concept relates to a first principle-derived effective mass approximation simulation technology for accurate and efficient computational simulation of an optical property of a quantum nanostructure including creating an effective mass approximation (EMA) parameter through first principle density functional theory (DFT) calculation for a model nanostructure corresponding to a simulation target quantum nanostructure, performing EMA calculation using the EMA parameter created through the DFT calculation, and acquiring the optical property of the quantum nanostructure based on an electronic structure generated through the EMA calculation.

Abstract: A processor-implemented method with image analysis includes: receiving a test image; generating a plurality of augmented images by augmenting the test image; determining classification prediction values for the augmented images using a classifier; determining a detection score based on the classification prediction values; and determining whether the test image corresponds to anomaly data based on the detection score and a threshold.

Abstract: In a method of responding based on sentence paraphrase recognition for dialog system, main keywords of a domain and patterns thereof are recognized and extracted based on a morpheme analysis result in a pre-processing process. Question domains/sub-categories/dialogue-acts are classified using the morpheme analysis result and the extracted main keywords and patterns. Learning a model is performed using classification features of the classified question domains, sub-categories, and dialogue-acts as semantic features of query sentences, and sentence semantic vectors are extracted by measuring similarity between the vectors. A language model of letters is trained and similarity in expression and structure is analyzed by extracting a sentence expression vector based on the letter.

Abstract: Disclosed are an ultra-high sensitivity target signal detection method based on analysis of a noise signal of a sensor using deep learning based anomaly detection and a system using the same. More particularly, disclosed are a method and apparatus for receiving a noise signal from a sensor and inputting data to an artificial neural network trained with a normal noise signal to determine whether or not a target signal is present. The target signal detection method is capable of detecting a target signal having a very low concentration that can be detected by a conventional sensor, whereby the target signal detection method is useful in developing an ultra-high sensitivity sensor.

Abstract: An MRAM cell includes a switch unit configured to determine opening and closing thereof by a word line voltage and to activate a current path between a bit line and a bit line bar in an opened state thereof, first and second MTJs having opposite states, respectively, and connected in series between the bit line and the bit line bar, to constitute a storage node, and a sensing line configured to be activated in a reading mode of the MRAM cell, thereby creating data reading information based on a voltage between the first and second MTJs, wherein the first and second MTJs have different ones of a low resistance state and a high resistance state, respectively, in accordance with a voltage drop direction between the bit line and the bit line bar, thereby storing data of 0 or 1.

Abstract: Disclosed are a method of manufacturing a carbon support for a fuel cell catalyst, a carbon support for a fuel cell catalyst manufactured according to the method, and a catalyst for a fuel cell including the same. The method may include using various organic materials containing N and various carbon supports and thus provide excellent economic feasibility. In addition, pyridinic N and pyrrolic N of doped N can be adjusted at an optimal content ratio so that the carbon support for a fuel cell catalyst manufactured and the catalyst for a fuel cell including the same have excellent electrochemical resistance and excellent electrochemical characteristic due to an increase in an electrochemically active surface area, and excellent durability due to an increase in thermal durability.