Osong Medical Innovation Foundation, Korea University Research and Business Foundation

Samsung Electronics Co., Ltd.

Skhynix Inc.

Korea Electronics Technology Institute

Dong Yeob Shin

Jong Sun Park

Geon Ho Kim

Joong Ho Jo

Jongsun Park

Jin Ho Jeong

Won Gyu Shin

Tae Beom Lim

Yong Seok Lim

Ho June Leu

Hoon Jo

Yeong Min Lee

Joong Sik Kim

The present invention is related to the electronic device and learning method for learning of low complexity artificial intelligence model based on selecting dynamic prediction confidence threshold.

Electronic device and learning method for learning of low complexity artificial intelligence model based on selecting dynamic prediction confidence threshold

A deep neural network training accelerator includes an operational unit sequentially performing first and second operations on a plurality of input data of a sub-set according to a mini-batch gradient descent, a determination unit determining each of the input data as one of skip data and training data based on a confidence matrix obtained by the first operation, and a control unit controlling the operational unit to skip the second operation with respect to the skip data.

Deep neural network training accelerator and operation method thereof

The present invention relates to a pop count-based deep learning neural network computation method, a multiply accumulator, and a device thereof. The computation method according to an exemplary embodiment of the present invention is a computation method for a deep learning neural network, including a step of generating one-hot encoding codes according to the type of first multiplication result values for a multiplication (first multiplication) of weights (W) and input values (A); a step of performing a pop-count for each generated code; and a step of accumulating result values for a constant multiplication (second multiplication) between each type of the first multiplication result value and each count value of the pop-count which are different constant values.

Pop count-based deep learning neural network computation method, multiply accumulator and device

An image processing device is provided. The image processing device includes an image signal processor for processing a raw image received from a camera and a memory for storing a previous frame of the raw image and an intermediate image generated by the processing, wherein the image signal processor estimates an initial global motion vector between the previous frame and a current frame, receives focus region information in the raw image, divides the raw image into a foreground (FG) region and a background (BG) region based on the focus region information, generates a final global motion vector by divisionally updating the initial global motion vector based on the FG region and the BG region, performs motion compensation by applying the final global motion vector to the previous frame, and outputs a final image by blending the motion-compensated previous frame with the current frame.

Image signal processor and image processing device

Disclosed is an erasure-based Reed-Solomon code soft-decision decoding method and device, capable of reducing a decoding time while minimizing the effect on error correction performance. The Reed-Solomon code soft-decision decoding device includes an erasure control circuit configured to determine whether a number of errors in a codeword is odd or even, and to provide a key equation solver circuit with a first erasure pattern or a second erasure pattern according to a result of the determining when a decoding failure is detected by a decoding error detection circuit, the first erasure pattern being provided when the number of errors is odd, the second erasure pattern being provided when the number of errors is even.

Reed-Solomon code soft-decision decoding method and device

An operation device of a convolutional neural network includes a first operation unit performing a convolution operation on an input feature map based on a predetermined mosaic type to output at least one pixel information of an output feature map, a prediction unit predicting the other pixel information of the output feature map as one of a skip pixel and an operation pixel based on the at least one pixel information, and a second operation unit skipping the convolution operation with respect to the sip pixel to process the skip pixel as an output of zero.

Growing community of inventors

Seoul, South Korea

Dong Yeob Shin