Abstract: Computer-aided methods for performing computations using adversarially-robust neural networks is disclosed that includes providing a uncoded neural network comprising a plurality of neurons and associated inputs for each of the plurality of neurons, each neuron configured to perform a calculation according to an activation function. The method further includes transforming, using the computing device, at least one uncoded neuron into a coded neuron by adding, using the computing device, an error correcting code as an additional input to the at least one uncoded neuron, the error correcting code comprising a redundant combination of the associated inputs off the uncoded neuron, and revising, using the computing device, the activation function of the at least one uncoded neuron to accommodate the error correcting code as the additional input.

Abstract: A radiation hardened NAND flash memory data storage device suitable for space flight having a plurality of memory cells configured to store data values in accordance with a predetermined rank modulation scheme and a memory controller that receives a current error count from an error decoder of the data device for one or more data operations of the flash memory device and selects an operating mode for data scrubbing in accordance with the received error count and a program cycles count. Methods of operating the data storage device are also described.

Abstract: A mutation profile can be determined for an individual's DNA sequence or sequence segment that provides information about the evolutionary history of the DNA. This mutation profile can then be used with a machine learning classifier trained on other people's mutation profiles to determine probabilities that the individual has certain phenotypes. An example is cancer, where the probabilities of different types of cancer can be provided in a disease risk propensity.

Abstract: Methods for distributed storage in accordance with embodiments of the invention enable secret sharing. One embodiment includes encoding source data using an encoding system to produce a plurality of sets of encoded data, where: the source data can be recovered from at least a portion of less than all of the plurality of sets of encoded data; and the source data cannot be recovered using less than a threshold number of the plurality of sets of encoded data; storing each of the plurality of sets of encoded data on a storage device from a set of storage devices on which encoded data is stored; determining a set of storage devices that are available using a decoding system, where the set of storage devices that are available does not include all of the storage devices in the set of storage devices on which encoded data is stored.

Abstract: Disclosed are low-complexity schemes to store information in a distributed manner across multiple nodes of a computer system. The schemes are resilient to node failures and resistant to node eavesdropping. The disclosed schemes may be used in conjunction with RAID systems, and when used in such systems may be referred to as“secure RAID”.

Abstract: Techniques are disclosed for generating codes for representation of data in memory devices that may avoid the block erasure operation in changing data values. Data values comprising binary digits (bits) can be encoded and decoded using the generated codes, referred to as codewords, such that the codewords may comprise a block erasure-avoiding code, in which the binary digits of a data message m can be encoded such that the encoded data message can be stored into multiple memory cells of a data device and, once a memory cell value is changed from a first logic value to a second logic value, the value of the memory cell may remain at the second logic value, regardless of subsequently received messages, until a block erasure operation on the memory cell.

Abstract: A radiation hardened NAND flash memory data storage device suitable for space flight having a plurality of memory cells configured to store data values in accordance with a predetermined rank modulation scheme and a memory controller that receives a current error count from an error decoder of the data device for one or more data operations of the flash memory device and selects an operating mode for data scrubbing in accordance with the received error count and a program cycles count. Methods of operating the data storage device are also described.

Abstract: A data device includes a memory having a plurality of memory cells configured to store data values in accordance with a predetermined rank modulation scheme that is optional and a memory controller that receives a current error count from an error decoder of the data device for one or more data operations of the flash memory device and selects an operating mode for data scrubbing in accordance with the received error count and a program cycles count.

Abstract: Disclosed are constructions of WOM codes that combine rewriting and error correction for mitigating the reliability and the endurance problems typically experienced with flash memory. A rewriting model is considered that is of practical interest to flash memory applications where only the second write uses WOM codes. The disclosed WOM code construction is based on binary erasure quantization with LDGM codes, where the rewriting uses message passing and has potential to share the efficient hardware implementations with LDPC codes in practice. The coding scheme achieves the capacity of the rewriting model.

Abstract: A data device includes a memory having a plurality of memory cells configured to store data values in accordance with a predetermined rank modulation scheme that is optional and a memory controller that receives a current error count from an error decoder of the data device for one or more data operations of the flash memory device and selects an operating mode for data scrubbing in accordance with the received error count and a program cycles count.

Abstract: The reliability of NAND flash memory decreases rapidly as density increases, preventing the wide adoptions of flash-based storage systems. A novel data representation scheme named rank modulation (RM) is discussed for improving NAND flash reliability. RM encodes data using the relative orders of memory cell voltages, which is inherently resilient to asymmetric errors. For studying the effectiveness of RM in flash, RM is adapted to make it simple to implement with existing flash memories. The implementation is evaluated under different types of noise of 20 nm flash memory. Results show that RM offers significantly lower cell error rates compared to the current data representation in flash at typical P/E cycles. RM is applied to flash-based archival storage and shows that RM brings up to six times longer data retention time for 16 nm flash memory.

Abstract: Both rewriting and error correction are technologies usable for non-volatile memories, such as flash memories. A coding scheme is disclosed herein that combines rewriting and error correction for the write-once memory model. In some embodiments, code construction is based on polar codes, and supports any number of rewrites and corrects a substantial number of errors. The code may be analyzed for a binary symmetric channel. The results can be extended to multi-level cells and more general noise models.

Abstract: Rank modulation has been recently proposed as a scheme for storing information in flash memories. Three improved aspects are disclosed. In one aspect the minimum push-up scheme, for storing data in flash memories is provided. It aims at minimizing the cost of changing the state of the memory. In another aspect, multi-cells, used for storing data in flash memories is provided. Each transistor is replaced with a multi-cell of mm transistors connected in parallel. In yet another aspect, multi-permutations, are provided. The paradigm of representing information with permutations is generalized to the case where the number of cells in each level is a constant greater than one. In yet another aspect, rank-modulation rewriting schemes which take advantage of polar codes, are provided for use with flash memory.

Abstract: A data device includes a memory having a plurality of memory cells configured to store data values in accordance with a predetermined rank modulation scheme that is optional and a memory controller that receives a current error count from an error decoder of the data device for one or more data operations of the flash memory device and selects an operating mode for data scrubbing in accordance with the received error count and a program cycles count.

Abstract: Disclosed are low-complexity schemes to store information in a distributed manner across multiple nodes of a computer system. The schemes are resilient to node failures and resistant to node eavesdropping. The disclosed schemes may be used in conjunction with RAID systems, and when used in such systems may be referred to as“secure RAID”.

Abstract: Rank modulation has been recently proposed as a scheme for storing information in flash memories. Three improved aspects are disclosed. In one aspect the minimum push-up scheme, for storing data in flash memories is provided. It aims at minimizing the cost of changing the state of the memory. In another aspect, multi-cells, used for storing data in flash memories is provided. Each transistor is replaced with a multi-cell of m transistors connected in parallel. In yet another aspect, multi-permutations, are provided. The paradigm of representing information with permutations is generalized to the case where the number of cells in each level is a constant greater than one.

Abstract: A data device includes a memory having a plurality of memory cells configured to store data values in accordance with a predetermined rank modulation scheme that is optional and a memory controller that receives a current error count from an error decoder of the data device for one or more data operations of the flash memory device and selects an operating mode for data scrubbing in accordance with the received error count and a program cycles count.

Abstract: Methods for distributed storage in accordance with embodiments of the invention enable secret sharing. One embodiment includes encoding source data using an encoding system to produce a plurality of sets of encoded data, where: the source data can be recovered from at least a portion of less than all of the plurality of sets of encoded data; and the source data cannot be recovered using less than a threshold number of the plurality of sets of encoded data; storing each of the plurality of sets of encoded data on a storage device from a set of storage devices on which encoded data is stored; determining a set of storage devices that are available using a decoding system, where the set of storage devices that are available does not include all of the storage devices in the set of storage devices on which encoded data is stored.

Abstract: Techniques are disclosed for generating codes for representation of data in memory devices that may avoid the block erasure operation in changing data values. Data values comprising binary digits (bits) can be encoded and decoded using the generated codes, referred to as codewords, such that the codewords may comprise a block erasure-avoiding code, in which the binary digits of a data message m can be encoded such that the encoded data message can be stored into multiple memory cells of a data device and, once a memory cell value is changed from a first logic value to a second logic value, the value of the memory cell may remain at the second logic value, regardless of subsequently received messages, until a block erasure operation on the memory cell.

Abstract: Disclosed are constructions of WOM codes that combine rewriting and error correction for mitigating the reliability and the endurance problems typically experienced with flash memory. A rewriting model is considered that is of practical interest to flash memory applications where only the second write uses WOM codes. The disclosed WOM code construction is based on binary erasure quantization with LDGM codes, where the rewriting uses message passing and has potential to share the efficient hardware implementations with LDPC codes in practice. The coding scheme achieves the capacity of the rewriting model.