Abstract: An example method is provided in according with one implementation of the present disclosure. The method comprises computing, via a processor, a ranked elements list for each of a plurality of objects. The method also comprises iteratively computing, via the processor, a blacklist of elements for the objects. The method further comprises determining, via the processor, duster centers that include top ranked non-blacklisted elements, and assigning, via the processor, each object to at least one duster center.

Abstract: An example method is provided in according with one implementation of the present disclosure. The method comprises generating, via a processor, a set of hashes for each of a plurality of objects. The method also comprises computing, via the processor, a high-dimensional sparse vector for each object, where the vector represents the set of hashes for each object. The method further comprises computing, via the processor, a combined high-dimensional sparse vector from the high-dimensional sparse vectors for all objects and computing a hash suppression threshold. The method also comprises determining, via the processor, a group of hashes to be suppressed by using the hash suppression threshold, and suppressing, via the processor, the group of selected hashes when performing an action.

Abstract: Disclosed is an improved systems, methods, and computer program products that use a cluster-based probability model to perform anomaly detection, where the clusters are based upon entities and interactions that exist in content management platforms.

Abstract: A system may include an access engine to access an input vector as well as a projection matrix. The projection matrix may include a number of rows equal to a number of hash values to generate from the input vector multiplied by the square root of an inverted sparsity parameter specifying a ratio of the hash universe size from which the hash values are generated to the number of hash values to generate. The projection matrix may include a number of columns equal to the dimensionality of the input vector. The system may also include a hash computation engine to determine a projection vector from the projection matrix and the input vector, split the projection vector into a number of sub-vectors equal to the number of hash values to generate, and generate a hash value from each of the sub-vectors.

Abstract: Incremental automatic update of ranked neighbor lists based on k-th nearest neighbors is disclosed. One example is a system including an indexing module to retrieve an incoming data stream, and retrieve ranked neighbor lists for received data objects. An evaluator determines similarity measures between the received data objects and their respective k-th nearest neighbors. A threshold determination module determines a statistical distribution based on the determined similarity measures, and a threshold based on the statistical distribution. The evaluator determines additional similarity measures between a new data object in the data stream and the received data objects.

Abstract: A system may include an access engine to access an input vector as well as a projection matrix. The projection matrix may include a number of rows equal to a number of hash values to generate from the input vector multiplied by the square root of an inverted sparsity parameter specifying a ratio of the hash universe size from which the hash values are generated to the number of hash values to generate. The projection matrix may include a number of columns equal to the dimensionality of the input vector. The system may also include a hash computation engine to determine a projection vector from the projection matrix and the input vector, split the projection vector into a number of sub-vectors equal to the number of hash values to generate, and generate a hash value from each of the sub-vectors.

Abstract: A system may include an access engine and a projection engine. The access engine may access a feature vector with an initial dimensionality that represents a data object of a physical system. The projection engine may generate an extended vector with an extended dimensionality from the feature vector. The projection engine may also apply an orthogonal transformation to the extended vector to obtain an intermediate vector with the extended dimensionality, as well as compute the inner products of the intermediate vector and sparse binary vectors of a sparse binary vector set. In doing so, the projection engine may obtain a randomly projected vector with an output dimensionality that is greater than the extended dimensionality of the intermediate vector. Then, the projection engine may output the randomly projected vector as an output vector that is a random projection of the feature vector with the output dimensionality.

Abstract: In some examples, a method includes accessing input vectors in an input space, wherein the input vectors characterize elements of a physical system. The method may also include generating feature vectors from the input vectors, and the feature vectors are generated without any vector product operations between performed between any of the input vectors. An inner product of a pair of the feature vectors may correlate to an implicit kernel for the pair of feature vectors, and the implicit kernel may approximate a Gaussian kernel within a difference threshold. The method may further include providing the feature vectors to an application engine for use in analyzing the elements of the physical system, other elements in the physical system, or a combination of both.

Abstract: In some examples, a system includes an access engine and a hyperplane determination engine. The access engine may access a training vector set that includes sparse binary training vectors and a set of labels classifying each of the sparse binary training vectors through a positive label or a negative label. The hyperplane determination engine may initialize a candidate hyperplane vector and maintain a scoring vector including scoring vector elements to track separation variances of the sparse binary training vectors with respect to the candidate hyperplane vector. Through iterations of identifying, according to the scoring vector, a particular sparse binary training vector with a greatest separation variance with respect to the candidate hyperplane vector, the hyperplane determination engine may incrementally update the candidate hyperplane vector and incrementally update the scoring vector to adjust separation variances affected by updates to the candidate hyperplane vector.

Abstract: Examples disclosed herein involve data stream analytics. In examples herein, a data stream may be analyzed by computing a set of hashes of a real-valued vector, the real-valued vector corresponding to a sample data object of a data stream; generating a list of data objects from a database corresponding to the sample data object based on the set of hashes, the list of data objects ordered based on similarity of the data objects to the sample data object of the data stream; and updating a data structure representative of activity of the sample data object in the data stream based on the list of data objects, the data structure to provide incremental analysis corresponding to the sample data object.

Abstract: A method includes receiving information about a plurality of data chunks and determining if one or more of a plurality of back-end nodes already stores more than a threshold amount of the plurality of data chunks where one of the plurality of back-end nodes is designated as a sticky node. The method further includes, responsive to determining that none of the plurality of back-end nodes already stores more than a threshold amount of the plurality of data chunks, deduplicating the plurality of data chunks against the back-end node designated as the sticky node. Finally, the method includes, responsive to an amount of data being processed, designating a different back-end node as the sticky node.

Abstract: In one embodiment, for a first image, a first vector of similarity to a set of reference images is calculated as a first face descriptor, and for a second image, a second vector of similarity to the set of reference images is calculated as a second face descriptor. A similarity measure between the first face descriptor and the second face descriptor is then calculated.

Abstract: Embodiments of the present invention relate to a system and method of document retention with policy-controlled deletion. Embodiments of the present invention comprise committing to a plurality of documents, deleting one of the plurality of documents, and providing a proof of authorized deletion of the one of the plurality of documents in response to an audit request.

Abstract: A plurality of differential data stores are stored in persistent storage media. In response to receiving a first request to store a particular data object, one of the differential data stores that are stored in the persistent storage media is selected, wherein selecting the one differential data store is according to a criterion relating to compression of data objects in the differential data stores. The selected differential data store is copied into temporary storage media, where the copying is not delayed after receiving the first request to await receipt of more requests. The particular data object is inserted into the copy of the selected differential data store in the temporary storage media, where the inserting is performed without having to retrieve more data from the selected differential store in the persistent storage media.

Abstract: To detect duplicative hierarchically arranged sets of files in a storage system, a method includes generating, for hierarchically arranged plural sets of files, respective collections of values computed based on files in corresponding sets of files. For a further set of files that is an ancestor of at least one of the plural sets of files, a respective collection of values that is based on the collection of values computed for the at least one set is generated. Duplicative sets according to comparisons of the collections of values are identified.

Abstract: Provided are, among other things, systems, methods and techniques for scanning information across multiple different devices. In one representative system, remote data-processing devices are provided with scanning applications that repeatedly scan information on their respective data-processing devices to identify matching data units that satisfy a specified matching criterion, the specified matching criterion including required matches against a set of screening digests, and then transmit characteristic information regarding the matching data units; and a central processing facility receives the characteristic information from the remote data-processing devices and determines whether the corresponding matching data units satisfy a policy criterion.

Abstract: A sequence of hashes is received. Each hash corresponds to a data chunk of data to be deduplicated. Locations of previously stored copies of the data chunks are determined, the locations determined based on the hashes. A breakpoint in the sequence of data chunks is determined based on the locations, the breakpoint forming a boundary of a segment of data chunks.

Abstract: Provided are, among other things, systems, methods and techniques for identifying related objects in a computer database. In one representative implementation: (a) a feature vector that describes an existing object is obtained; (b) comparison scores are generated between the feature vector and various sample vectors; (c) a set that includes at least one designated vector is identified from among the sample vectors by evaluating the generated comparison scores; (d) a computer database is searched for matches between label(s) for the designated vector(s) and labels for representative vectors for other objects represented in the computer database; and (e) at least one related object is identified based on the identified match(es).

Abstract: A method includes receiving information about a plurality of data chunks and determining if one or more of a plurality of back-end nodes already stores more than a threshold amount of the plurality of data chunks where one of the plurality of back-end nodes is designated as a sticky node. The method further includes, responsive to determining that none of the plurality of back-end nodes already stores more than a threshold amount of the plurality of data chunks, deduplicating the plurality of data chunks against the back-end node designated as the sticky node. Finally, the method includes, responsive to an amount of data being processed, designating a different back-end node as the sticky node.

Abstract: A method for data deduplication includes receiving a set of hashes derived from a data chunk of a set of input data chunks 310. The method includes sampling the set of hashes 320, using an index indentifying data chunk containers that hold data chunks having a hash in the set of sampled hashes 330, and loading indexes for at least one of the identified data chunk containers 340. The method includes determining which of the hashes correspond to data chunks stored in data chunk containers corresponding to the loaded indexes 350 and deciding which of the set of input data chunks should be stored based at least in part on the determination.