Abstract: A method for training a sequence transduction model includes receiving a sequence of unlabeled input features extracted from unlabeled input samples. Using a teacher branch of an unsupervised subnetwork, the method includes processing the sequence of input features to predict probability distributions over possible teacher branch output labels, sampling one or more sequences of teacher branch output labels, and determining a sequence of pseudo output labels based on the one or more sequences of teacher branch output labels. Using a student branch that includes a student encoder of the unsupervised subnetwork, the method includes processing the sequence of input 10 features to predict probability distributions over possible student branch output labels, determining a negative log likelihood term based on the predicted probability distributions over possible student branch output labels and the sequence of pseudo output labels, and updating parameters of the student encoder.

Abstract: Proposed is an apparatus for performing high-capacity wireless communication in a greenhouse environment, the apparatus including: a moving enabling control unit including a motor for moving at least one wireless router in an upper end portion of a greenhouse, a moving enabling module including a motor control module and a motor communication module, and a moving enabling rails installed in a predetermined number, length and direction at the upper end portion of the greenhouse, and a greenhouse integrated control unit configured to set an interval of movement of the wireless router by communicating with the motor communication module and to generate and transmit a control command for controlling the motor control module in such a manner that the wireless router moves at the set interval of movement.

Abstract: The present invention relates to a method, a device and a computer-readable medium for training a machine learning model performing competency evaluation on a plurality of competencies, and more particularly, to a method, a device and a computer-readable medium for training a machine learning model performing competency evaluation on a plurality of competencies to efficiently train the machine learning model for outputting an evaluation result of each of the competencies from input data related to answers and the like of an evaluatee.

Abstract: A heat exchanger includes a heat transfer pipe for guiding a refrigerant; and a plurality of fins which respectively have a through-hole through which the heat transfer pipe passes and are disposed spaced apart from each other to allow air to pass in a first direction, wherein the fin includes: a corrugated form portion having an inclination with the first direction which is an air flow direction; and a sheet portion comprising a surface parallel to the first direction around the through hole, the sheet portion has a first length in the first direction which is an air flow direction, and has a second length longer than the first length in a second direction perpendicular to the air flow direction.

Abstract: A method includes receiving a plurality of unlabeled audio samples corresponding to spoken utterances not paired with corresponding transcriptions. At a target branch of a contrastive Siamese network, the method also includes generating a sequence of encoder outputs for the plurality of unlabeled audio samples and modifying time characteristics of the encoder outputs to generate a sequence of target branch outputs. At an augmentation branch of a contrastive Siamese network, the method also includes performing augmentation on the unlabeled audio samples, generating a sequence of augmented encoder outputs for the augmented unlabeled audio samples, and generating predictions of the sequence of target branch outputs generated at the target branch. The method also includes determining an unsupervised loss term based on target branch outputs and predictions of the sequence of target branch outputs. The method also includes updating parameters of the audio encoder based on the unsupervised loss term.

Abstract: A robot according to an embodiment of the present invention comprises: at least one wheel for traveling; a water tank in which an accommodating space accommodating water is formed; a water pump connected to the water tank; a water purifier connected to the water pump to purify and discharge water provided from the water tank; and an air cleaner configured to suction and filter air and discharge the filtered air, wherein the water tank is disposed below the water pump, the water purifier, and the air cleaner.

Abstract: The present invention relates to an anti-glare film including: a light-transmitting substrate; and a hard coating layer containing a binder resin and inorganic fine particles dispersed in the binder resin, wherein the anti-glare film has a specific reflection intensity ratio (R), and a polarizing plate and a display device including the anti-glare film.

Abstract: A display device includes a substrate including an active area and an inactive area which encloses the active area, a common voltage line which is disposed in the inactive area and is connected to the light emitting diode, an encapsulation unit which covers the active area and a part of the inactive area, a touch sensing unit disposed on the encapsulation unit, and an auxiliary line which is disposed on the encapsulation unit and is connected to the common voltage line, the touch sensing unit includes a plurality of touch electrodes and a bridge pattern which connects the plurality of second touch electrodes, and the auxiliary line is formed of the same material as the bridge pattern and is disposed on the same layer as the bridge pattern, and the auxiliary line is disposed in both the active area and the inactive area.

Abstract: A terahertz device includes a terahertz element, a sealing resin, a wiring layer and a frame-shaped member. The terahertz element that performs conversion between terahertz waves and electric energy. The terahertz element has an element front surface and an element back surface spaced apart from each other in a first direction. The sealing resin covers the terahertz element. The wiring layer is electrically connected to the terahertz element. A frame-shaped member is made of a conductive material and arranged around the terahertz element as viewed in the first direction. The frame-shaped member has a reflective surface capable of reflecting the terahertz waves.

Abstract: Disclosed are a method and an apparatus for converting a part of a dental image. The method for converting a part of a dental image may comprise the steps of: receiving a dental image; receiving a user input corresponding to a conversion region of the dental image and a target type; and generating an output dental image in which the conversion region has been converted into the target type by using a pre-trained generation model.

Abstract: The present invention relates to a heat exchanger comprising multiple arranged fin tubes, each of which comprises a heat transferring fin and multiple tubes integrally formed on a plate, wherein refrigerant flows through the tubes. Each of the fin tubes has a first surface forming the front surface of the fin tube and a second surface forming the rear surface of the fin tube, and has opening part which communicates with at least a side of the multiple tubes and is formed to extend through the first surface and the second surface; and the perimeters of the openings of the fin tubes adjacent to each other may be connected through a connection part such that the openings of the adjacent fin tubes communicate with each other. Therefore, there is an advantage in that refrigerant can be distributed without a separate header.