Abstract: A data storage system including a memory array including a plurality of memory devices programmable in greater than two states. A memory control module may control operations of the memory array, and an encoder module may encode input data for storing to the memory array. The memory array may be an m×n memory array, and the memory control module may control operations of storing data to and retrieving data from the memory array.

Abstract: A method is provided to manage access to a memory array. The method includes encoding a bit string with a rank metric encoder to generate an encoded binary array, modifying the encoded binary array so each row has at most half of the row with a bit value and each column has at most half of the column with the bit value, and storing the modified binary array into corresponding memory devices of the memory array.

Abstract: A programmable crossbar array with inline fuses includes a layer of row lines and a layer of column lines with the row lines crossing over the column lines to form junctions and resistive memory elements sandwiched between row lines and a column lines at the junctions. Inline fuses are located in either the row lines, column lines or both. The inline fuses are interposed between the support circuitry and the resistive memory elements. A method for mitigating shorts in a crossbar array is also provided.

Abstract: A method for rewriting a memory array with a number of memory elements includes performing a rewrite process to change the memory array from an initial state to a target state in a manner that avoids violating a set of weight constraints at any time during the rewrite process. A memory system includes a memory array and a memory controller configured to perform a rewrite process to change the memory array from an initial state to a target state in a manner that avoids violating a set of weight constraints at any time during the rewrite process.

Abstract: One example disclosed in the application is an electronic data-storage device comprising one or more arrays of memory elements that each includes a data-storage medium that is switched between two different states by application of a switching-inducing force or gradient to the data-storage medium, a top control element and a bottom control element through which the switching-inducing force or gradient is applied, and a feedback signal. The data-storage device also includes an error-control-coding encoder that encodes received data and a READ/WRITE controller that writes encoded data received from the error-control-coding encoder to a number of memory elements by applying the switching-inducing force to the one or mare arrays of memory elements until feedback signals indicate that the WRITE operation has completed or until the switching-inducing force or gradient has been applied for a maximum application time.

Abstract: One embodiment of the present invention is directed to an electronic memory comprising a memory element and control logic that determines a data state of the memory element by: measuring a physical characteristic of the memory element to obtain a measured value corresponding to the data state, measuring the physical characteristic of the memory element, after setting the memory element to a first known data state, to obtain a first calibration value, comparing the measured value corresponding to the data state with the first calibration value to determine the data state; and determining a reliability metric for the determined data state.

Abstract: One example disclosed in the application is an electronic data-storage device comprising one or more arrays of memory elements. The data-storage device also includes an error-control-coding encoder that encodes received data and a READ/WRITE controller that writes encoded data received from the error-control-coding encoder to a number of memory elements by applying the switching-inducing force or gradient to the one or more arrays of memory elements until more than a maximum-allowed number of WRITE requests have been queued to the WRITE-request buffer, until feedback signals indicate that the WRITE operation has completed, or until the switching-inducing force or gradient has been applied for a maximum application time.

Abstract: One embodiment of the present invention is directed to a method for compressing data generated by multiple data sources. The method includes steps of partitioning data generated by the multiple data sources into data partitions, the data included in each data partition containing inter-data-source redundancies and, for each data partition, compressing the data in the data partition to remove the inter-data-source redundancies.

Abstract: A multi-state memory system with encoding that minimizes half-select currents. The system includes an array of row and column conductors with a plurality of storage cells each capable of being placed into any of three or more physical states. An encoder is connected to receive data bits for storage and to apply activation signals to the row and column conductors to write information to the storage cells. The encoder is programmed to encode the data bits into entries in an array having one row corresponding with each row conductor and one column corresponding with each column conductor; select entries in the array according to half-select currents of the storage cells; apply a predetermined one-dimensional mapping that increases the value of at most one entry in the array to obtain a mapped array; and write entries of the mapped array into the storage cells.

Abstract: One embodiment of the present invention is directed to an electronic memory (102) comprising a memory element (104) and control logic that determines a data state of the memory element by: measuring (702) a physical characteristic of the memory element to obtain a measured value corresponding to the data state, measuring (708) the physical characteristic of the memory element, after setting (706) the memory element to a first known data state, to obtain a first calibration value, comparing the measured value corresponding to the data state with the first calibration value to determine the data state (714); and determining a reliability metric (716) for the determined data state.

Abstract: A data storage system including a memory array including a plurality of memory devices programmable in greater than two states. A read/write control module may overwrite data in the memory array without violating a constraint during the overwrite process. The memory array may be an m×n memory array.

Abstract: A technology is provided for compressing digital discrete node data to reduce overall power consumption. Node data can be represented by a plurality of data units with a specified data width and can also be viewed as a plurality of bit planes corresponding to data at each bit position for the data units. A threshold bit position value may be selected for data units using an achievable compressibility estimate relative to an estimated energy consumption. The threshold bit position value can represent a boundary where an estimated energy consumption for compressing and transmitting a bit plane is less than an estimated energy consumption for transmitting the bit plane uncompressed. A bit plane is selected in the plurality of bit planes with a bit position value greater than the threshold bit position value. The bit plane is compressed using a compressor in the networked node.

Abstract: A method for rewriting a memory array (408) with a number of memory elements (206) includes performing a rewrite process to change the memory array (408) from an initial state to a target state in a manner that avoids violating to a set of weight constraints at any time during the rewrite process. A memory system includes a memory array (408) and a memory controller (104) configured to perform a rewrite process to change the memory array (408) from an initial state to a target state in a manner that avoids violating a set of weight constraints at any time during the rewrite process.

Abstract: One embodiment of the present invention is directed to an adaptive context-based predictor that predicts a value {circumflex over (x)} from a context, stored in an electronic memory, corresponding to a noisy-dataset symbol zi of a noisy dataset corrupted with noise modeled as being introduced by a noise-introducing channel. The adaptive context-based predictor is adapted according to one or more parameters that specify adaptive context-based-predictor operation, at least one of which functionally depends, or partially functionally depends, on a level of noise represented by the noise-introducing channel. The adaptive context-based predictor computes a number of intermediate values from the context, computes the predicted value {circumflex over (x)} from the intermediate values, and stores the predicted value {circumflex over (x)} in the electronic memory.

Abstract: A data storage system including a memory array including a plurality of memory devices programmable in greater than two states. A memory control module may control operations of the memory array, and an encoder module may encode input data for storing to the memory array. The memory array may be an m×n memory array, and the memory control module may control operations of storing data to and retrieving data from the memory array.

Abstract: A data storage system including a memory array including a plurality of memory devices programmable in greater than two states. A read/write control module may overwrite data in the memory array without violating a constraint during the overwrite process. The memory array may be an m×n memory array.

Abstract: A method of storing data into a memory array converts an input string into a first binary array with (m?1) rows and (n?1) columns. A second binary array with m rows and n columns in an encoded bit pattern is then generated from the first binary array. The second binary array in the encoded bit pattern has at most n/2 1's in each row and at most m/2 1's in each column, and the m-th row and an n-th column contain information for decoding other entries of the second binary array. The encoded bit pattern of the second binary array is then stored into corresponding memory devices of the memory array.

Abstract: Embodiments of the present invention are directed to various enhanced discrete-universal denoisers that have been developed to denoise images and other one-dimensional, two-dimensional or higher-dimensional data sets in which the frequency of occurrence of individual contexts may be too low to gather efficient statistical data or context-based symbol prediction. In these denoisers, image quality, signal-to-noise ratios, or other measures of the effectiveness of denoising that would be expected to increase monotonically over a series of iterations may decrease, due to assumptions underlying the discrete-universal-denoising method losing validity. Embodiments of the present invention apply context-class-based statistics and statistical analysis to determine, on a per-context-class basis, when to at least temporarily terminate denoising iterations on each conditioning class.

Abstract: A random number generator that generates random numbers based on measurements of one or more environmental quantities. The random number generator includes a first sensor for generating a first sequence of digital values representing measurements of a first environmental quantity at successive times. A first compressor that provides a first sequence of compressed values having a lower internal correlation than the values of the first sequence of digital values compresses the sequence of digital values. The random number generator generates an output random number via a circuit for generating a random number from an input sequence of digital values, the input sequence being a function of one of the first sequence of compressed values. The circuit may utilize a hash function to provide further security. In addition, the circuit includes a blocking circuit for preventing the generator from outputting a random number if the input sequences fail a predetermined test.

Abstract: Embodiments of the present invention provide context-class-based universal denoising of noisy images and other noise-corrupted data sets. Prediction-error statistics for each prediction class, relative to a prefiltered image, are collected to estimate a bias for each prediction class, and prediction-error statistics for each conditioning class, relative to a prefiltered image, are accumulated based on the difference between predicted values and corresponding prefiltered-image symbols. The prediction-error statistics are accumulated using computed prediction-error-statistics vectors, with inversion of a prediction-error vector generated from each prediction prior to accumulation in a prediction-error-statistics vector. Conditional probability distributions are computed for individual contexts, which allow for computing a clean-image-estimated, value for each noisy-image value by minimizing a computed distortion over a range of possible estimated-clean-image symbols.