Abstract: A composite vector is received. A first candidate component vector is generated and evaluated. The first candidate component vector is selected, based on the evaluating, as an accurate component vector. The first candidate component vector is unbundled from the composite vector. The unbundling results in a first reduced vector.

Abstract: A probability distribution corresponding to the kernel function is determined and weights are sampled from the determined probability distribution corresponding to the given kernel function. Memristive devices of an analog crossbar are programmed based on the sampled weights, where each memristive device of the analog crossbar is configured to represent a corresponding weight. Two matrix-vector multiplication operations are performed on an analog input x and an analog input y using the programmed crossbar and a dot product is computed on results of the matrix-vector multiplication operations.

Abstract: Photonic content-addressable memories (CAMs) and applications thereof are provided. The CAM includes a photonic cross-bar array comprising a plurality of row and column waveguides, and a plurality of photonic filter devices. Each filter device is selectively programmable in first and second states representing respective stored bit values that filters out light according to the programming. An encoder for encoding a plurality of input bit-strings into optical signals such that bit values in different bit-strings are encoded using optical signals in different pairs of optical states, and to simultaneously supply the optical signals corresponding to each bit-position in the bit-strings to a respective row waveguide of the array. The CAM further comprises a detector for detecting light in any of said optical states in each column waveguide, thereby identifying any mismatch between each input bit-string and bit values stored in the filter devices coupling light to that waveguide.

Abstract: The invention is notably directed to a device for performing a matrix-vector multiplication of a matrix with a vector. The device comprises a memory crossbar array comprising a plurality of row lines, a plurality of column lines and a plurality of junctions arranged between the plurality of row lines and the plurality of column lines. Each junction comprises a programmable resistive element and an access element for accessing the programmable resistive element. The device further comprises a readout circuit configured to perform read operations by applying positive read voltages of one or more first amplitudes and negative read voltages of one or more second amplitudes corresponding to the one or more first amplitudes. The one or more first amplitudes and the corresponding one or more second amplitudes are different from each other, thereby correcting polarity dependent current asymmetricities.

Abstract: Methods and apparatus are provided for memorizing data signals in a spiking neural network. For each data signal, such a method includes supplying metadata relating to the data signal to a machine learning model trained to generate an output signal, indicating a relevance class for a data signal, from input metadata for that data signal. The method includes iteratively supplying the data signal to a sub-assembly of neurons, interconnected via synaptic weights, of a spiking neural network and training the synaptic weights to memorize the data signal in the sub-assembly. The method further comprises assigning neurons of the network to the sub-assembly in dependence on the output signal of the model such that more relevant data signals are memorized by larger sub-assemblies. The data signal memorized by a sub-assembly can be subsequently recalled by activating neurons of that sub-assembly.

Abstract: A 3D compute-in-memory accelerator system and method for efficient inference of Mixture of Expert (MoE) neural network models. The system includes a plurality of compute-in-memory cores, each in-memory core including multiple tiers of in-memory compute cells. One or more tiers of in-memory compute cells correspond to an expert sub-model of the MoE model. One or more expert sub-models are selected for activation propagation based on a function-based routing, the tiers of the corresponding experts being activated based on this function. In one embodiment, this function is a hash-based tier selection function used for dynamic routing of inputs and output activations. In embodiments, the function is applied to select a single expert or multiple experts with input data-based or with layer-activation-based MoEs for single tier activation. Further, the system is configured as a multi-model system with single expert model selection or with a multi-model system with multi-expert selection.

Abstract: The disclosure includes a computer-implemented method of factorizing a vector by utilizing resonator network modules. Such modules include an unbinding module, as well as search-in-superposition modules. The method includes the following steps. A product vector is fed to the unbinding module to obtain unbound vectors. The latter represent estimates of codevectors of the product vector. A first operation is performed on the unbound vectors to obtain quasi-orthogonal vectors. The first operation is reversible. The quasi-orthogonal vectors are fed to the search-in-superposition modules, which rely on a single codebook. In this way, transformed vectors are obtained, utilizing a single codebook. A second operation is performed on the transformed vectors. The second operation is an inverse operation of the first operation, which makes it possible to obtain refined estimates of the codevectors.

Abstract: A computerized neuro-vector-symbolic architecture, that: receives image data associated with an artificial intelligence (AI) task; processes the image data using a frontend that comprises an artificial neural network (ANN) and a vector-symbolic architecture (VSA); and processes an output of the frontend using a backend that comprises a symbolic logical reasoning engine, to solve the AI task. The AI task, for example, may be an abstract visual reasoning task.

Abstract: The invention is notably directed to a device comprising a plurality of resistive memory elements. The plurality of resistive memory elements comprises a resistive material. The device is configured to apply programming pulses to a subset of the plurality of resistive memory elements to perform a temporary resistance change of the resistive material of the subset for a predefined retention period, thereby programming the subset of the plurality of resistive elements from a first resistance state corresponding to a first binary state to a second resistance state corresponding to a second binary state. The device is configured such that a resistance of the subset of the plurality of resistive elements reverts automatically during the predefined retention period from the second resistance state to the first resistance state by an inherent material property of the resistive material, thereby automatically deleting the second binary state.

Abstract: Embodiments are disclosed for a method. The method includes determining a granularity of hypervectors. The method also includes receiving an input hypervector representing a data structure. Additionally, the method includes performing an iterative process to factorize the input hypervector into individual hypervectors representing the cognitive concepts. The iterative process includes, for each concept: determining an unbound version of a hypervector representing the concept by a blockwise unbinding operation between the input hypervector and estimate hypervectors of other concepts. The iterative process further includes determining a similarity vector indicating a similarity of the unbound version of the hypervector with each candidate code hypervector of the concept. Additionally, the iterative process includes generating an estimate of a hypervector representing the concept by a linear combination of the candidate code hypervectors, and weights of the similarity vector.

Abstract: A computer-implemented method for managing datasets of a storage system is provided, wherein the datasets have respective sets of metadata, the method including: successively feeding first sets of metadata to a spiking neural network (SNN), the first sets of metadata fed corresponding to datasets of the storage system that are labeled with respect to classes they belong to, so as to be associated with class labels, for the SNN to learn representations of said classes in terms of connection weights that weight the metadata fed; successively feeding second sets of metadata to the SNN, the second sets of metadata corresponding to unlabeled datasets of the storage system, for the SNN to infer class labels for the unlabeled datasets, based on the second sets of metadata fed and the representations learned; and managing datasets in the storage system, based on class labels of the datasets, these including the inferred class labels.

Abstract: The invention is notably directed to a device for performing a matrix-vector multiplication of a matrix with a vector. The device comprises a memory crossbar array comprising a plurality of row lines, a plurality of column lines and a plurality of junctions arranged between the plurality of row lines and the plurality of column lines. Each junction comprises a programmable resistive element and an access element for accessing the programmable resistive element. The device further comprises a readout circuit configured to perform read operations by applying positive read voltages of one or more first amplitudes and negative read voltages of one or more second amplitudes corresponding to the one or more first amplitudes. The one or more first amplitudes and the corresponding one or more second amplitudes are different from each other, thereby correcting polarity dependent current asymmetricities.

Abstract: Provided is a method, device, and computer program product for programming a set of first elements onto a computational memory. The computational memory allows for performing a computation task from a set of second elements that encode the set of first elements in the computational memory, respectively. The method includes performing the computation task by the computational memory using the set of second elements and adapting at least part of the set of second elements in the computational memory based on a measured result of the computation task, until the measured result of the computation task fulfils an accuracy condition.

Abstract: The invention is notably directed to a method of processing data in-memory. The method applies electrical signals to at least two input lines, which correspond to at least two rows. These two rows include at least one of the K rows and at least one of the L rows. This causes to obtain output signals in output of the M output lines, wherein the output signals depend on target values and operand values, in accordance with data stored across said at least two rows. Finally, the output signals are read out and a transformation operation is concurrently performed, in-memory, on the target values based on the operand values. This way transformed data are obtained by way of in-memory processing. The transformation may for instance be a cryptographic operation; the operand data may encode a cryptographic key. The invention is further directed to related apparatuses and systems, notably cryptographic service systems.

Abstract: Predefined concepts are represented by codebooks. Each codebook includes candidate code hypervectors that represent items of a respective concept of the predefined concepts. A neuromorphic memory device with a crossbar array structure includes row lines and column lines stores a value of respective code hypervectors of an codebook. An input hypervector is stored in an input buffer. A plurality of estimate buffers are each associated with a different subset of row lines and a different codebook and initially store estimated hypervectors. An unbound hypervector is computed using the input hypervector and all the estimated hypervectors. An attention vector is computed that indicates a similarity of the unbound hypervector with one estimated hypervector. A linear combination of the one estimated hypervector, weighted by the attention vector, is computed and is stored in the estimate buffer that is associated with the one estimated hypervector.

Abstract: A computer device, a non-transitory computer storage medium, and a computer-implemented method of pattern recognition utilizing an elastic clustering algorithm. A sequence of input datapoints are assigned to a particular cluster of K clusters based on a distance from a centroid k representing a center of the particular cluster. The centroid k in each of the K clusters is shifted from a first position to a second position closer than the first position from the sequence of input datapoints. A location of the centroid k in each of the K clusters is relaxed from the second position toward an equilibrium point in the particular cluster of the K clusters. The relaxing of the location of the centroid k occurs according to an elasticity pull factor based on a distance between the centroid k of the particular cluster at a time t.

Abstract: A computer-implemented method for performing a classification of an input signal utilizing a neural network includes: computing, by a feature extraction unit of the neural network, a query vector; and performing, by a classification unit, a factorization of the query vector to a plurality of codebook vectors of a plurality of codebooks to determine a corresponding class of a number of classes. A set of combinations of vector products of the plurality of codebook vectors of the plurality of codebooks establishes a number of classes of the classification unit.

Abstract: A computer-implemented method for performing a classification of an input signal by a neural network includes: computing, by a feature extraction unit of the neural network, a D-dimensional query vector, wherein D is an integer; generating, by a classification unit of the neural network, a set of C fixed D-dimensional quasi-orthogonal bipolar vectors as a fixed classification matrix, wherein C is an integer corresponding to a number of classes of the classification unit; and performing a classification of a query vector based, at least in part, on the fixed classification matrix.

Abstract: A computer-implemented method for factorizing hypervectors in a resonator network includes: receiving an input hypervector representing a data structure; performing an iterative process for each concept in a set of concepts associated with the data structure in order to factorize the input hypervector into a plurality of individual hypervectors representing the set of concepts, wherein the iterative process includes: generating a first estimate of an individual hypervector representing a concept in the set of concepts; generating a similarity vector indicating a similarity of the estimate of the individual hypervector with each candidate attribute hypervector of a plurality of candidate attribute hypervectors representing an attribute associated with the concept; and generating a second estimate of the individual hypervector based, at least in part, on a linear combination of the plurality of candidate attribute hypervectors and performing a non-linear function on the linear combination of the plurality of ca

Abstract: The invention is directed to solving an optimization problem. The method operates a photonic crossbar array structure including N input lines and M output lines, which are interconnected at junctions via N×M photonic memory devices, where N?2 and M?2. The photonic memory devices are programmed to store respective weights in accordance with the optimization problem. The photonic crossbar array structure is operated as follows. First, the method determines values of L input vectors of N components each, where L?2. Second, based on the determined values, N electromagnetic signals are generated, where each of the generated signals multiplexes L input signals encoded at respective wavelengths, so as for the N electromagnetic signals to map the L input vectors of N components each. Third, the N electromagnetic signals generated are applied to the N input lines of the photonic crossbar array structure.