Abstract: An embodiment of the present disclosure discloses a memory device. The memory device comprises a memory controller, a buffer and a memory array. The buffer is coupled to the memory controller or embedded in the memory controller. A storage space of the buffer is configured by the memory controller to include a plurality of groups. The memory array is coupled to the memory controller, and comprising a plurality of tiles. The groups are one-to-one corresponding to the tiles. Each of the groups is configured to store data to be written into the corresponding tile. The memory controller performs one or more write operations based on the groups.

Abstract: A memory device, includes a memory array for storing a plurality of vector data each of which has an MSB vector and a LSB vector. The memory array includes a plurality of memory units each of which has a first bit and a second bit. The first bit is used to store the MSB vector of each vector data, the second bit is used to store the LSB vector of each vector data. Each vector data is executed with a multiplying-operation, the MSB vector and the LSB vector of each vector data is executed with a first group-counting operation and a second group-counting operation respectively. The threshold voltage distribution of each memory unit is divided into N states, where N is a positive integer and N is less than 2 to the power of 2, the effective bit number stored by each memory unit is less than 2.

Abstract: A memory device, includes a memory array for storing a plurality of vector data each of which has an MSB vector and a LSB vector. The memory array includes a plurality of memory units each of which has a first bit and a second bit. The first bit is used to store the MSB vector of each vector data, the second bit is used to store the LSB vector of each vector data. A bit line corresponding to each vector data executes one time of bit-line-setup, and reads the MSB vector and the LSB vector of each vector data according to the bit line. The threshold voltage distribution of each memory unit is divided into N states, where N is a positive integer and N is less than 2 to the power of 2, and the effective bit number stored by each memory unit is less than 2.

Abstract: Disclosed is a memory device and an operation method thereof. The operation method of memory device, comprising: programming a plurality of sub-matrices including at least one of non-zero element of a rearranged matrix to a plurality of operation units of the memory device; and programming a mapping table into a working memory of a memory device. The rearranged matrix is generated by rearrange a plurality of columns and a plurality of rows of an original matrix according to the positions of a plurality of non-zero elements of the original matrix. The mapping table comprises a correspondence of row indexes between the original matrix and the rearranged matrix, a correspondence of column indexes between the original matrix and the rearranged matrix and a correspondence between the sub-matrices including at least one non-zero element and the operation units storing the sub-matrices including at least one non-zero element.

Abstract: A method for neural network computation using adaptive data representation, adapted for a processor to perform multiply-and-accumulate operations on a memory having a crossbar architecture, is provided. The memory comprises multiple input and output lines crossing each other, multiple cells respectively disposed at intersections of the input and output lines, and multiple sense amplifiers respectively connected to the output lines. In the method, an input cycle of kth bits respectively in an input data is adaptively divided into multiple sub-cycles, wherein a number of the divided sub-cycles is determined according to a value of k. The kth bits of the input data are inputted to the input lines with the sub-cycles and computation results of the output lines are sensed by the sense amplifiers. The computation results sensed in each sub-cycle are combined to obtain the output data corresponding to the kth bits of the input data.

Abstract: A memory device includes: a memory array used for implementing neural networks (NN); and a controller coupled to the memory array. The controller is configured for: in updating and writing unrewritable data into the memory array in a training phase, marching the unrewritable data into a buffer zone of the memory array; and in updating and writing rewritable data into the memory array in the training phase, marching the rewritable data by skipping the buffer zone.

Abstract: A memory device includes: a memory array used for implementing neural networks (NN), the NN including a plurality of layers; and a controller coupled to the memory array, the controller being configured for: determining a computation duration of a first data of a first layer of the plurality of layers; selecting a first program operation if the computation duration of the first data of the first layer is shorter than a threshold; and selecting a second program operation if the computation duration of the first data of the first layer is longer than the threshold, wherein the second program operation has a longer program pulse time than the first program operation.

Abstract: A method for neural network computation using adaptive data representation, adapted for a processor to perform multiply-and-accumulate operations on a memory having a crossbar architecture, is provided. The memory comprises multiple input and output lines crossing each other, multiple cells respectively disposed at intersections of the input and output lines, and multiple sense amplifiers respectively connected to the output lines. In the method, an input cycle of kth bits respectively in an input data is adaptively divided into multiple sub-cycles, wherein a number of the divided sub-cycles is determined according to a value of k. The kth bits of the input data are inputted to the input lines with the sub-cycles and computation results of the output lines are sensed by the sense amplifiers. The computation results sensed in each sub-cycle are combined to obtain the output data corresponding to the kth bits of the input data.

Abstract: A method and an apparatus for neural network computation using adaptive data representation, adapted for a processor to perform multiply-and-accumulate operations on a memory having a crossbar architecture, are provided. The memory comprises multiple input and output lines crossing each other, multiple cells respectively disposed at intersections of the input and output lines, and multiple sense amplifiers respectively connected to the output lines. In the method, an input cycle of kth bits respectively in an input data is adaptively divided into multiple sub-cycles, wherein a number of the divided sub-cycles is determined according to a value of k. The kth bits of the input data are inputted to the input lines with the sub-cycles and computation results of the output lines are sensed by the sense amplifiers. The computation results sensed in each sub-cycle are combined to obtain the output data corresponding to the kth bits of the input data.

Abstract: A memory page management method is provided. The method includes receiving a state-change notification corresponding to a state-change page, and grouping the state-change page from a list to which the state-change page belongs into a keep list or an adaptive LRU list of an adaptive adjusting list according to the state-change notification; receiving an access command from a CPU to perform an access operation to target page data corresponding to a target page; determining that a cache hit state is a hit state or a miss state according to a target NVM page address corresponding to the target page, and grouping the target page into the adaptive LRU list according to the cache hit state; and searching the adaptive page list according to the target NVM page address to obtain a target DRAM page address to complete the access command corresponding to the target page data.

Abstract: A computing in memory method for a memory device is provided. The computing in memory method includes: based on a stride parameter, unfolding a kernel into a plurality of sub-kernels and a plurality of complement sub-kernels; based on the sub-kernels and the complement sub-kernels, writing a plurality of weights into a plurality of target memory cells of a memory array of the memory device; inputting an input data into a selected word line of the memory array; performing a stride operation in the memory array; temporarily storing a plurality of partial sums; and summing the stored partial sums into a stride operation result when all operation cycles are completed.

Abstract: Disclosed is a memory device and an operation method thereof. The operation method of memory device, comprising: programming a plurality of sub-matrices including at least one of non-zero element of a rearranged matrix to a plurality of operation units of the memory device; and programming a mapping table into a working memory of a memory device. The rearranged matrix is generated by rearrange a plurality of columns and a plurality of rows of an original matrix according to the positions of a plurality of non-zero elements of the original matrix. The mapping table comprises a correspondence of row indexes between the original matrix and the rearranged matrix, a correspondence of column indexes between the original matrix and the rearranged matrix and a correspondence between the sub-matrices including at least one non-zero element and the operation units storing the sub-matrices including at least one non-zero element.

Abstract: A computer-implemented method and a system for configuring and implementing an artificial neural network. The method for configuring includes initializing an artificial neural network; and training, based on a training operation, the artificial neural network so as to form an adaptively deployable artificial neural network defining a plurality of nested artificial neural sub-networks. Each artificial neural sub-network is optimal for a respective resource configuration. The method for implementing includes determining an optimal configuration of the artificial neural network for deployment at an electrical device, and deploying the artificial neural network with the optimal configuration at the electrical device.

Abstract: The present disclosure provides a memory system, a method of operating memory, and a non-transitory computer readable storage medium. The memory system includes a memory chip and a controller. The controller is coupled with the memory chip, which the controller is configured to: receive a first data corresponding to a first version from a file system in order to store the first data corresponding to the first version in a first page of the flash memory chip; and program the first data corresponding to a second version in the first page in response to the first data of the second version, which the second version is newer than the first version.

Abstract: An in-memory computing method and apparatus, adapted for a processor to perform MAC operations on a memory, are provided. In the method, a format of binary data of weights is transformed from a floating-point format into a quantized format by truncating at least a portion of fraction bits of the binary data and calculating complements of remaining bits, and programming the transformed binary data into cells of the memory. A tuning procedure is performed by iteratively inputting binary data of input signals into the memory, integrating outputs of the memory, and adjusting the weights programmed to the cells based on the integrated outputs. The binary data of the weights is reshaped based on a probability of reducing bits with a value of one in the binary data of each weight. The tuning procedure is repeated until an end condition is met.

Abstract: A memory management method includes: performing a bloom filtering operation on a plurality of logic block addresses to determine a read and written frequency of each of the logic block addresses; setting a first program/erase (P/E) cycle threshold and a second P/E cycle threshold value, wherein the first P/E cycle threshold value is smaller than the second P/E cycle threshold value; dividing each of physical memory blocks into a first type memory block, a second type memory block or a third type memory block according to the first P/E cycle threshold value and the second P/E cycle threshold value; and, allocating each of the logic block addresses to the first type memory block, the second type memory block or the third type memory block according to the read and written frequency of corresponding logic block addresses.

Abstract: A computing in memory method for a memory device is provided. The computing in memory method includes: based on a stride parameter, unfolding a kernel into a plurality of sub-kernels and a plurality of complement sub-kernels; based on the sub-kernels and the complement sub-kernels, writing a plurality of weights into a plurality of target memory cells of a memory array of the memory device; inputting an input data into a selected word line of the memory array; performing a stride operation in the memory array; temporarily storing a plurality of partial sums; and summing the stored partial sums into a stride operation result when all operation cycles are completed.

Abstract: The present disclosure provides a memory system, a method of operating memory, and a non-transitory computer readable storage medium. The memory system includes a memory chip and a controller. The controller is coupled with the memory chip, which the controller is configured to: receive a first data corresponding to a first version from a file system in order to store the first data corresponding to the first version in a first page of the flash memory chip; and program the first data corresponding to a second version in the first page in response to the first data of the second version, which the second version is newer than the first version.

Abstract: A memory management method includes: performing a bloom filtering operation on a plurality of logic block addresses to determine a read and written frequency of each of the logic block addresses; setting a first program/erase (P/E) cycle threshold and a second P/E cycle threshold value, wherein the first P/E cycle threshold value is smaller than the second P/E cycle threshold value; dividing each of physical memory blocks into a first type memory block, a second type memory block or a third type memory block according to the first P/E cycle threshold value and the second P/E cycle threshold value; and, allocating each of the logic block addresses to the first type memory block, the second type memory block or the third type memory block according to the read and written frequency of corresponding logic block addresses.

Abstract: A memory device includes: a memory array used for implementing neural networks (NN), the NN including a plurality of layers; and a controller coupled to the memory array, the controller being configured for: determining a computation duration of a first data of a first layer of the plurality of layers; selecting a first program operation if the computation duration of the first data of the first layer is shorter than a threshold; and selecting a second program operation if the computation duration of the first data of the first layer is longer than the threshold, wherein the second program operation has a longer program pulse time than the first program operation.