Abstract: An Out-Of-Distribution (OOD) detection method performed by an OOD detection device is provided. The OOD detection method includes generating a noise filter using training data based on a target distribution and using a classification model pre-trained according to the target distribution, training the generated noise filter to reduce a loss function for a first output value obtained by applying the training data passed through the generated noise filter to the pre-trained classification model, and detecting Out-Of-Distribution (OOD) for new data using a second output value obtained by applying the new data passed through the trained noise filter to the pre-trained classification model.

Abstract: A side outer extraction method may include receiving, by at least one processor, an image, of a vehicle, preprocessed from three-dimensional (3D) data from a computer-aided design (CAD) module, detecting, using an artificial intelligence model, a classification value and a bounding box for each region, of a plurality of regions, corresponding to one of a plurality of target references of the preprocessed image, transmitting, to the CAD module, a signal indicating the classification value and the bounding box for each region of the plurality of regions, and causing extraction, by the CAD module, of the plurality of target references from the classification value and the bounding box for each region of the plurality of regions, based on the received signal.

Abstract: The present disclosure relates to a method and an apparatus for federated learning of an artificial intelligence model. According to an exemplary embodiment of the present disclosure, a federated learning method of an artificial intelligence model includes: training a first local artificial intelligence model and a second local artificial intelligence model using data sets of a first client and a second client among the plurality of clients; calculating performance values for the first local artificial intelligence model and the second local artificial intelligence model by transmitting the first local artificial intelligence model to the second client and transmitting the second local artificial intelligence model to the first client; comparing the performance values to remove one of the first client and the second client; and training a global model using a client which is not removed.

Abstract: An Out-Of-Distribution (OOD) detection method performed by an OOD detection device is provided. The OOD detection method includes generating a noise filter using training data based on a target distribution and using a classification model pre-trained according to the target distribution, training the generated noise filter to reduce a loss function for a first output value obtained by applying the training data passed through the generated noise filter to the pre-trained classification model, and detecting Out-Of-Distribution (OOD) for new data using a second output value obtained by applying the new data passed through the trained noise filter to the pre-trained classification model.

Abstract: A local model training method of a federated learning framework implementing training data classification is provided. In the local model training method, a client may classify training data into two categories, generate a learning mini-batch by adjusting a ratio between samples classified into the two categories and included in the mini-batch to a preset ratio, and train a learning model using the mini-batch with the adjusted sample ratio.

Abstract: A neural network model training method and apparatus are provided. The method includes a first training operation of training the neural network model with original data, the first training operation including generating a first feature map for the original data, and generating a first class activation map for the original data from the generated first feature map, and a second training operation of training the neural network model with adversarial data transformed from the original data, the second training operation including generating a second feature map for the adversarial data, generating a second class activation map for the adversarial data from the generated second feature map, and training the neural network model so that the second class activation map follows the first class activation map based on logit pairing for the first and second class activation maps.

Abstract: Provided is a method of removing, by a server, an abnormal client in federated learning. A method of removing, by a server, an abnormal client in federated learning may include receiving, from a user equipment (UE), first weight values trained in a first local model, generating a first client model based on the first weight values, validating the first client model by using a validation data set in order to determine whether the first client model is legitimate, and removing the first weight values based on the first client model not being legitimate.

Abstract: Provided are a method and an apparatus for generating a keyword. The method for generating a keyword in a target document includes: extracting primitive single words from the target document; generating a candidate keyword by a combination of the primitive single words; calculating cohesion score between a first word and a second word included in the candidate keyword; calculating a context score of the first word and the second word based on similarity between two contexts by determining a periphery of a word where the first word and the second word simultaneously occur in a neighboring document as one context and determining the target document as a remaining one context; and selecting a final keyword based on the cohesion score between the first word and the second word and the context score of the first word and the second word.

Abstract: Provided are a method and an apparatus for pruning based on the number of updates, the method comprising counting the number of updates of weights for each epoch and removing n weights for which the number of updates is small after training of a network is completed, wherein n represents the number of weights to be removed to satisfy a desired sparsity.

Abstract: Provided are a method and an apparatus for generating a keyword. The method for generating a keyword in a target document includes: extracting primitive single words from the target document; generating a candidate keyword by a combination of the primitive single words; calculating cohesion score between a first word and a second word included in the candidate keyword; calculating a context score of the first word and the second word based on similarity between two contexts by determining a periphery of a word where the first word and the second word simultaneously occur in a neighboring document as one context and determining the target document as a remaining one context; and selecting a final keyword based on the cohesion score between the first word and the second word and the context score of the first word and the second word.

Abstract: The present invention relates to an inner-plasticized vinyl chloride-based copolymer resin not requiring plasticizers and a preparation method thereof. Specifically, the vinyl chloride-based copolymer resin is prepared by a suspension polymerization method of initiating the polymerization of vinyl chloride monomer, feeding a certain amount of butyl acrylate continuously or discontinuously thereinto and carrying out the additional polymerization at the temperature higher than the polymerization initiation temperature so as to prepare a core-shell type vinyl chloride-based random copolymer resin. The vinyl chloride-based copolymer resin of core-shell structure prepared by the present invention includes vinyl chloride-butyl acrylate copolymer, and it can provide a vinyl chloride-butyl acrylate copolymer product which can be processed without plasticizers positively necessary to produce a soft product.

Abstract: The present invention relates to an inner-plasticized vinyl chloride-based copolymer resin not requiring plasticizers and a preparation method thereof. Specifically, the vinyl chloride-based copolymer resin is prepared by a suspension polymerization method of initiating the polymerization of vinyl chloride monomer, feeding a certain amount of butyl acrylate continuously or discontinuously thereinto and carrying out the additional polymerization at the temperature higher than the polymerization initiation temperature so as to prepare a core-shell type vinyl chloride-based random copolymer resin. The vinyl chloride-based copolymer resin of core-shell structure prepared by the present invention includes vinyl chloride-butyl acrylate copolymer, and it can provide a vinyl chloride-butyl acrylate copolymer product which can be processed without plasticizers positively necessary to produce a soft product.

Abstract: A system for controlling abnormal traffic based on a fuzzy logic includes: an intrusion detection module for analyzing packets incoming from a network interface by means of a membership function defined based on a specific period of time, and outputting a fuzzy value representing a degree of a port scan attack; a fuzzy control module for recognizing the degree of the port scan attack based on the fuzzy value and outputting a control signal for traffic control according to the recognized degree of the port scan attack; and an intrusion blocking module for receiving the control signal and controlling the traffic with the network interface.

Abstract: Provided are an ontology system, a method for managing the ontology system, and a recording medium storing the same. The ontology system includes: a context broker unit for receiving context information from a sensing device and verifying a validity of the received context information; a context managing unit for controlling to generate an ontology structure by transforming the verified context information from the context broker unit to ontology web language (OWL) data and processing the OWL data; a rule-based inference engine unit for transforming the processed context information from the context managing unit to semantic web rule language (SWRL) data and processing the SWRL data through an inference process; a learning managing unit for processing the processed context information from the context managing unit through learning; and a database for storing the context information processed at the context managing unit, the rule-based engine unit, and the learning managing unit.

Abstract: Disclosed is a sequence similarity measuring apparatus and a method of controlling the same. The sequence similarity measuring apparatus using dynamic programming includes: a matrix generating unit for generating a matrix based on the dynamic programming by using two sequences; a normalization unit for calculating a similarity reference value by inputting an element value of a last row/column of the matrix generated by the matrix generating unit into a normalization formula for a given sequence length; and a similarity measuring unit for measuring predefined sequence similarity between the two sequences, based on the similarity reference value calculated by the normalization unit. This makes it possible to easily and correctly achieve similarity comparison between multiple sequences, and thus this technology is expected to be widely utilized in biology/programming application fields.

Abstract: A system for controlling abnormal traffic based on a fuzzy logic includes: an intrusion detection module for analyzing packets incoming from a network interface by means of a membership function defined based on a specific period of time, and outputting a fuzzy value representing a degree of a port scan attack; a fuzzy control module for recognizing the degree of the port scan attack based on the fuzzy value and outputting a control signal for traffic control according to the recognized degree of the port scan attack; and an intrusion blocking module for receiving the control signal and controlling the traffic with the network interface.

Abstract: Provided are an ontology system, a method for managing the ontology system, and a recording medium storing the same. The ontology system includes: a context broker unit for receiving context information from a sensing device and verifying a validity of the received context information; a context managing unit for controlling to generate an ontology structure by transforming the verified context information from the context broker unit to ontology web language (OWL) data and processing the OWL data; a rule-based inference engine unit for transforming the processed context information from the context managing unit to semantic web rule language (SWRL) data and processing the SWRL data through an inference process; a learning managing unit for processing the processed context information from the context managing unit through learning; and a database for storing the context information processed at the context managing unit, the rule-based engine unit, and the learning managing unit.