Abstract: An AI neural network accelerator for a transformer neural network includes a calculator configured to predict output tokens in units of input tokens and including a plurality of transformer operation cores operating based on a transformer model using n weights (n being a natural number), a weight generator including a weight embedding logic generated in advance as a result of training by matching weights of the transformer neural network with a×b kernel location information of the transformer neural network, and configured to generate an implicit weight based on location information of a kernel input from an external memory, and a controller configured to control an operation of each of the transformer operation cores after determining a size of the transformer model based on a number of weights.

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 method of quantifying fluorescent dyes labeled on extracellular vesicles, including providing an extracellular vesicle molecule labeled with a fluorescence dye molecule, and determining a brightness of each of the extracellular vesicle molecule labeled with the fluorescence dye molecule and the fluorescence dye molecule through fluorescence correlation spectroscopy (FCS) of the extracellular vesicle molecule and the fluorescence dye molecule.

Abstract: The present disclosure relates to a method of providing information for predicting a treatment response to an immune checkpoint inhibitor in a cancer patient by using multiplex immunohistochemistry, wherein, by performing multiplex immunohistochemistry on tumor tissue of a cancer patient to measure an expression level of an immune checkpoint molecule by an automated method, the treatment response to the immune checkpoint inhibitor in the cancer patient can be accurately and quickly predicted. In addition, unlike existing methods using single immunohistochemistry, the disclosed method can reduce errors of an inspector by analyzing markers simultaneously expressed in a single cell and evaluating the same by an automated method, and thus will be widely used as a companion diagnostic method for an immune checkpoint inhibitor.

Abstract: A device for evaluating characteristics of a target ground containing a metal component is proposed. The device includes a penetration probe having a main frame and a pair of side frames respectively installed at opposite side ends of the main frame, wherein each side frame has a lower end thereof extending downward from the main frame; a plurality of electrodes installed to be exposed to outside on the main frame; an electrode measurement part for measuring apparent chargeability of the target ground by applying power for measurement to the plurality of electrodes; and a main processor for calculating a weight ratio of a metal component of the target ground on the basis of the apparent chargeability measured by the electrode measurement part and calculating a volume ratio of the metal component of the target ground on the basis of the calculated weight ratio of the metal component.

Abstract: A device for evaluating characteristics of a target ground containing a metal component is proposed. The device includes: a penetration probe having a main frame and a pair of side frames respectively installed at opposite side ends of the main frame, wherein each side frame has a lower end thereof extending downward from the main frame; a plurality of electrodes installed to be exposed to outside on the main frame; an electrode measurement part for measuring apparent chargeability of the target ground by applying power for measurement to the plurality of electrodes; and a main processor for calculating a weight ratio of a metal component of the target ground on the basis of the apparent chargeability measured by the electrode measurement part and calculating a volume ratio of the metal component of the target ground on the basis of the calculated weight ratio of the metal component.

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: There is disclosed a multi-wavelength endoscopic system for imaging an observation site labeled with a plurality of fluorescent materials having different colors. The system includes an imaging unit configured to acquire image data by polarizing incident light reflected from the observation site in a first direction and a second direction perpendicular to the first direction, dividing a spectrum region of the incident light polarized in the first direction and the second direction into a plurality of spectrum channels and measuring the intensity of light for each of the spectrum channel. The system further includes a computing unit configured to store a single fluorescence spectrum extracted from sample image data obtained by single-treating the observation site with each of the fluorescent materials and configured to separate and output the image data obtained in the imaging unit using the single fluorescence spectrum so that each of the fluorescent materials is displayed separately.

Abstract: There is disclosed a multi-wavelength endoscopic system for imaging an observation site labeled with a plurality of fluorescent materials having different colors. The system includes an imaging unit configured to acquire image data by polarizing incident light reflected from the observation site in a first direction and a second direction perpendicular to the first direction, dividing a spectrum region of the incident light polarized in the first direction and the second direction into a plurality of spectrum channels and measuring the intensity of light for each of the spectrum channel. The system further includes a computing unit configured to store a single fluorescence spectrum extracted from sample image data obtained by single-treating the observation site with each of the fluorescent materials and configured to separate and output the image data obtained in the imaging unit using the single fluorescence spectrum so that each of the fluorescent materials is displayed separately.

Abstract: The present invention relates to a high-sensitivity liquid field-effect sensor for colon cancer, applicable to a sample such as blood or stool. The sensor according to one aspect enables ultra-high precision/low-concentration detection of colon cancer biomarkers, thereby having an effect of enabling early diagnosis of colon cancer even with a very small amount of a sample.