Korea Advanced Institute of Science and Technology

Samsung Electronics Co., Ltd.

Skhynix Inc.

Leesup Kim

Wongyu Shin

Yongkee Kwon

Youngsuk Moon

Jaemin Jang

Sehwan Lee

Hyeonuk Kim

Jaehyeong Sim

Yeongjae Choi

Donghyuk Kwon

Jaekang Shin

A learning method of an incremental learning model includes setting a searching range of mask weights based on either one or both of a distribution of mask weights of a binary mask corresponding to a filter of a pretrained model and a learning rate-related parameter, identifying a targeted mask weight in the searching range of the mask weights, updating the targeted mask weight based on the binary mask and the pretrained model, and updating a portion of the binary mask based on the updated targeted mask weight and a preset reference value.

Incremental learning system with selective weight updates

A memory system may include: a memory controller; a plurality of ranks; and a rank shared bus configured to couple the memory controller and the plurality of ranks. Each of the plurality of ranks may include: a plurality of banks; a rank bus coupled to the plurality of banks and configured to selectively transmit data to the rank shared bus or an intermediate buffer and selectively receive data from the rank shared bus or the intermediate buffer; and an intermediate buffer configured to be selectively coupled to the rank bus or the rank shared bus, according to a first signal from the memory controller.

Memory device supporting rank-level parallelism and memory system including the same

A processor-implemented neural network method includes: obtaining, from a memory, data of an input feature map and kernels having a binary-weight, wherein the kernels are to be processed in a layer of a neural network; decomposing each of the kernels into a first type sub-kernel reconstructed with weights of a same sign, and a second type sub-kernel for correcting a difference between a respective kernel, among the kernels, and the first type sub-kernel; performing a convolution operation by using the input feature map and the first type sub-kernels and the second type sub-kernels decomposed from each of the kernels; and obtaining an output feature map by combining results of the convolution operation.

Neural network method and apparatus

A processor-implemented neural network method includes: obtaining, from a memory, data an input feature map and kernels having a binary-weight, wherein the kernels are to be processed in a layer of a neural network; decomposing each of the kernels into a first type sub-kernel reconstructed with weights of a same sign, and a second type sub-kernel for correcting a difference between a respective kernel, among the kernels, and the first type sub-kernel; performing a convolution operation by using the input feature map and the first type sub-kernels and the second type sub-kernels decomposed from each of the kernels; and obtaining an output feature map by combining results of the convolution operation.

Growing community of inventors

Daejeon, South Korea

Leesup Kim