Abstract: In response to registration requests, metadata of a plurality of machine learning artifacts are stored. In response to a pipeline definition request, which does not specify resources to be used for the pipeline, a representation of a pipeline comprising nodes corresponding to registered artifacts is stored. A benchmarking operation is conducted to select the types of resources to be used for the nodes, and results of the benchmarking are provided.

Abstract: Hardly any work in literature attempts employing Function-as-a-Service (FaaS) or serverless architecture to accelerate the training or re-training process of meta-learning architectures. Embodiments of the present disclosure provide a method and system for meta learning using distributed training on serverless architecture. The system, interchangeably referred to as MetaFaaS, is a meta-learning based scalable architecture using serverless distributed setup. Hierarchical nature of gradient based architectures is leveraged to facilitate distributed training on the serverless architecture. Further, a compute-efficient architecture, efficient Adaptive Learning of hyperparameters for Fast Adaptation (eALFA) for meta-learning is provided. The serverless architecture based training of models during meta learning enables unlimited scalability and reduction of training time by using optimal number of serverless instances.

Abstract: A processor-implemented neural network method includes calculating individual update values for a weight assigned to a connection relationship between nodes included in a neural network; generating an accumulated update value by adding the individual update values; and training the neural network by updating the weight using the accumulated update value in response to the accumulated update value being equal to or greater than a threshold value, wherein the threshold value is a value of 2n of an n-th bit of the weight, where the n-th bit is a bit of lesser significance than a bit in the weight representing a largest magnitude bit among all bits of the weight.

Abstract: The present teaching relates to method, system, and computer programming product for dynamic vector allocation. Machine learning is conducted using training data constructed based on a target vector having a plurality of feature entries, wherein each of the plurality of feature entries is mapped from at least one original attribute from one or more original source vectors. A feature entry in the target vector is identified based on a first criterion associated with an assessment of the machine learning, for replacing the corresponding at least one original attribute from the one or more original source vectors. At least one alternative attribute from alternative source vectors based on a second criterion is determined, wherein the at least one alternative attribute is to be mapped to the feature entry of the target vector. The feature entry of the target vector is populated based on the at least one alternative attribute.

Abstract: A method for obtaining a user portrait includes: obtaining a user feature vector of a target user and tag feature vectors of content tags of multimedia content in a target application, and determining an alternative tag of the target user according to similarities between the user feature vector and the tag feature vectors, to further determine a user portrait of the target user according to the alternative tag.

Abstract: A computer-implemented device and corresponding method are provided for processing a multi-agent system input to form an at least partially optimised output indicative of an action policy. The method comprises receiving the multi-agent system input, the multi-agent system input comprising a definition of a multi-agent system and defining behaviour patterns of a plurality of agents based on system states; receiving an indication of an input system state; performing an iterative machine learning process to estimate a single aggregate function representing the behaviour patterns of the plurality of agents over a set of system states; and iteratively processing the single aggregate function for the input system state to estimate an at least partially optimised set of actions for each of the plurality of agents in the input system state. This may allow policies corresponding to the Nash equilibrium to be learned.

Abstract: One example method includes registering, by a customer, with a service provider, receiving, by the customer from the service provider, a global machine learning model, running, by the customer, the global machine learning model as a local machine learning model, collecting, by the customer, unlabeled data generated by edge devices operating in a customer domain, checking, by the customer, to determine if the customer domain has changed, and when it is determined that the customer domain has changed, performing, by the customer, a model adaptation process on the local machine learning model, and transmitting to the service provider, by the customer, gradients that comprise customer implemented changes to the local machine learning model.

Abstract: A method is provided for determining a control sequence for performing a multiqudit algorithm on a quantum computer, the multi-qudit algorithm expressible as a series of one or more k-qudit interactions. The method comprises, for each of the k-qudit interactions, decomposing the k-qudit interaction into a sequence of single-qudit unitary rotations and/or two-qudit unitary rotations from the continuous family of controllable unitary rotations generated by underlying physical interactions in the hardware of the quantum computer subject to a specified minimum interaction time, said sequence being physically implementable on the quantum computer. The method further comprises combining the sequences to form a combined interaction sequence. The method further comprises determining, based on the combined interaction sequence, the control sequence for performing the multiqudit algorithm on the quantum computer.

Abstract: A method of training a machine learning system for an object recognition device. The method includes: providing sensing element data; and training a machine learning system, using the provided sensing element data; at least one object being recognized from the sensing element data; and signal intensities of the sensing element data being used together with a reflection and/or absorption factor associated with the object.

Abstract: A quantum computing system includes a classical computer coupled in combination with a quantum computer, wherein the quantum computing system is configurable to execute program instructions to process input data to generate corresponding output data. The program instructions include one or more arithmetic functions to be executed using the quantum computer. The quantum computing system is configured to apply a transformation to transform the one or more arithmetic functions into a series of Fourier components that are executable using the quantum computer by using one or more quantum circuits utilizing rotation gates acting on qubits representing the Fourier components, and to process outputs from the one or more quantum circuits to generate results of the one or more arithmetic functions, wherein the results are used to generate the corresponding output data.

Abstract: A computation apparatus that includes a spiking neuron model. A spiking neuron model varies an index value of a signal output based on an input condition of a signal during an input time interval and outputs, based on the index value, a signal during an output time interval that starts after the input time interval ends.

Abstract: Various embodiments are provided for providing enhanced data allocation for machine learning operations in a computing environment by one or more processors in a computing system. One or more data sampling strategies may be determined based on a dataset. One or more enhanced training data allocations may be suggested for machine learning operations in a cloud computing environment based on the one or more data sampling strategies.

Abstract: Techniques for auto composing using a transformer-based language model having a parameter sharing decoder pair (PSDP) that reduces a number of parameters of the model and maintains the capability of generating understandable and reasonable compositions. In one particular aspect, a method is provided that includes obtaining a full encoder sequence, and inputting the full encoder sequence into a transformer model having a PSDP. The PSDP includes: a first decoder having parameters that are shared across all N layers of the first decoder; and a second decoder having parameters that are shared across all N layers of the first decoder. The parameters of the first decoder are different from the parameters of the second decoder. The method further includes using the transformer model to predict sequence elements based on the full encoder sequence, generate an output sequence comprising the sequence elements, and output an output sequence different from the full encoder sequence.

Abstract: Distributed quantum file consolidation is disclosed. A controlling quantum computing system (QCS) determines to consolidate a quantum file that includes a plurality of qubits implemented on a plurality of quantum computing systems (QCSs) onto a target QCS, the plurality of qubits including at least a first qubit implemented on a first QCS of the plurality of QCSs. The controlling QCS causes a transfer of quantum information contained in each qubit of the plurality of qubits that is not currently implemented on the target QCS to a corresponding qubit on the target QCS. Quantum file update information that indicates the qubits that compose the quantum file are located on the target QCS is communicated to at least the first QCS.

Abstract: Techniques are described for neural networks based on Progressive Neural ODEs (PODEs). In an example, a method to progressively train a neural ordinary differential equation (NODE) model comprises processing, by a machine learning system executed by a computing system, first training data, the first training data having a first complexity, to perform training of a first layer for the NODE model; and after performing the first training, processing second training data, the second training data having a second complexity that is higher than the first complexity, to perform training of a second layer for the NODE model.

Abstract: Methods, systems, and computer program products perform classification based on an imbalanced dataset. In a method, machine learning models are generated based on positive samples included in an imbalanced dataset. An amount of the positive samples is less than an amount of negative samples that are included in the imbalanced dataset. Each sample in the positive and negative samples includes parameters. Multiple influential parameter groups are identified from the parameters for the positive samples, respectively. A final predictive model is determined based on the machine learning models and the multiple influential parameter groups. The final predictive model is used for classifying a sample as a positive type or a negative type.

Abstract: Generally, the present disclosure provides systems and methods for performing machine learning in hyperbolic space. Specifically, techniques are provided which enable the learning of a classifier (e.g., large-margin classifier) for data defined within a hyperbolic space (e.g., which may be particularly beneficial for data possessing a hierarchical structure).

Abstract: A quantum computing system that supports efficient multitasking receives messages from a classical computing system to a pool of qubits. Each received message is associated with a partition identifier. The system configures a first set of qubits in the pool of qubits to perform a first computing task based on received messages that are associated with a first partition identifier and a second set of qubits in the pool of qubits to perform a second computing task based on received messages that are associated with a second partition identifier. The system acquires a first set of measurements from the first set of qubits and a second set of measurements from the second set of qubits. The system relays the first and second sets of measurements to the classical computing system.

Abstract: A method for predicting a molecular structure includes: preparing a learning data set including first learning data including an eigenvector value and a quantum mechanics calculation value for a monoatomic and molecular structural model, a bulk structural model, a slab structural model, and a nanoparticle structural model of a material including a plurality of elements; learning an artificial neural network using the learning data set to obtain a potential value; and predicting a molecular structure of another material by using the potential value.

Abstract: A method, an apparatus, and a system for configuring a neural network across heterogeneous processors are provided. The method includes creating a unified neural network profile for the plurality of processors; receiving at least one request to perform at least one task using the neural network; determining a type of the requested at least one task as one of an asynchronous task and a synchronous task; and parallelizing processing of the neural network across the plurality of processors to perform the requested at least one task, based on the type of the requested at least one task and the created unified neural network profile.