Abstract: Methods, systems, and apparatuses for adapting a predictive model for a manufacturing process. One method includes receiving, with an electronic processor, a plurality of data points for a plurality of manufactured parts and the predictive model. The predictive model outputs a label for a manufactured part provided by the manufacturing process indicating whether the manufactured part is accepted or rejected. The method also includes estimating, with the electronic processor, a drift for each of the plurality of data points and generating, with the electronic processor, an adapted version of the predictive model based on the predictive model and the drift for each of the plurality of data points. In addition, the method includes outputting, with the electronic processor, a label for each of the plurality of manufactured parts using the adapted version of the predictive model.

Abstract: In an example, for each of one or more terms in a text document, one or more entities to which the term potentially maps are identified. The text document includes at least one ambiguous term. One or more features are extracted from the text document. An attention model is applied to the text document based on the extracted one or more features, resulting in an attention weight being applied to each of the one or more terms in the text document. The one or more terms are encoded based on the attention weights. Each of one or more ambiguous terms is classified based on the encoded terms, the classification assigning a value to each different entity that each ambiguous term potentially maps to. A minimum entropy loss function is evaluated using the classification, and results are back-propagated to the attention model.

Abstract: An example system includes prediction workers, training workers, and a parameter server. The prediction workers store a local copy of a machine-learned model and run the mode exclusively in serving mode. The training workers store a local copy of a machine-learned model and a local snapshot and run the local copy exclusively in training mode and compare the local model or state to the snapshot after training to send delta updates to the parameter server after training. The parameter server aggregates received delta updates into a master copy of the model, sends the aggregated updates back to training workers and provides two types of updates; a real-time update based on a comparison of the master model with a local snapshot, and a full update. The real-time update occurs at least an order of magnitude more frequently than the full update and includes a subset of the weights in the model.

Abstract: Systems and methods for determining whether two tree persons in a genealogical database correspond to the same real-life individual. Embodiments include identifying two tree persons in a genealogical database and extracting a plurality of features from both tree persons to generate two vectors. Embodiments also include calculating a plurality of metrics between the two vectors to generate a metric function. Embodiments further include generating feature weights using a recursive process based on training data input by external users, and generating a score by calculating a weighted sum of the metric function being weighted by the feature weights. The generated score may then be compared to a threshold value.

Abstract: The present disclosure includes methods and systems for generating captions for digital images. In particular, the disclosed systems and methods can train an image encoder neural network and a sentence decoder neural network to generate a caption from an input digital image. For instance, in one or more embodiments, the disclosed systems and methods train an image encoder neural network (e.g., a character-level convolutional neural network) utilizing a semantic similarity constraint, training images, and training captions. Moreover, the disclosed systems and methods can train a sentence decoder neural network (e.g., a character-level recurrent neural network) utilizing training sentences and an adversarial classifier.

Abstract: A system and method for predicting remaining lifetime of a component of equipment is provided. The prediction system includes a data module, a feature module, a current data-based prediction module, a historical data-based prediction module, and a confidence module. The data module obtains a test sensor data of the component of equipment. The feature module obtains a historical health indicator and the current-health indicator. The current data-based prediction module obtains a first predicted remaining lifetime and a first prediction confidence according to the current-health indicator. The historical data-based prediction module obtains a second predicted remaining lifetime and a second prediction confidence according to the historical health indicator.

Abstract: Validating a task being performed on an HVAC system is described herein. One system includes a plurality of control devices associated with a heating, ventilation, and air conditioning (HVAC) system, wherein each respective control device is configured to record an action, taken by that control device as part of a task being performed on the HVAC system, as a block in a block chain for the task, send the block to the other control devices for validation of the action in the block chain for the task, update, upon the validation of the recorded action, the block chain for the task with the block having the recorded action, and store the updated block chain for the task.

Abstract: A method of machine learning includes performing dimensionality reduction on a parameter space by performing initial tests to determine scores for a plurality of parameter values in the parameter space, determining aggregate scores for a plurality of parameter value combinations, determining a ranking of the plurality of parameter value combinations based on the aggregate scores, and performing cluster analysis on the plurality of parameter value combinations to determine a set having highest aggregate scores. The method further includes performing additional tests, wherein each additional test is for a parameter value combination in the set. For each such parameter value combination, a probability of achieving a key performance indicator (KPI) is computed. Cluster analysis is then performed to determine a first subset of the set having highest probabilities of achieving the KPI. An operation is then performed on the first subset.

Abstract: Provided is a method for transmitting controlling control variables from a windfarm controller to units including at least one wind power installation or at least one energy store. The method include determining first and second controlling control variable components by the windfarm controller, outputting the first controlling control variable component in a first data packet, outputting the second controlling control variable component in a second data packet, receiving the first and second data packets by a first unit, and determining a controlling control variable from the first and second controlling control variable components. The first data packet has a receiver address which is assigned to the first unit and to at least one further unit, and the second data packet has a receiver address which is assigned to at least the first unit. Provided is a windfarm controller, a wind power installation and a windfarm configured to perform the method.

Abstract: Systems and methods are provided for interfacing multiple layers of optimization for a model of one or more processes in a processing environment to achieve increased or maximized stability in the underlying layer. To improve consistency between the solutions at the different model levels, the lower level of optimization can have extra constraints added to the optimization problem which target variables at their unconstrained values in the upper layer of optimization. The systems and methods can facilitate selection of variables to receive an external target such that stability of the solution is improved or maximized. This can be achieved, at least in part, by identifying variables that provide a reduced or minimized condition number for a sub-matrix in the lower level model when an additional external constraint is applied.

Abstract: Provided is a spiking neural network system for dynamical control of flexible, stable, and hybrid memory storage. An information storage method may include converting input information to a temporal pattern in a form of a spike; and storing the information that is converted to the temporal pattern in a spiking neural network. The storing may comprise storing information by applying, to the spiking neural network, a spike-timing-dependent plasticity (STDP) learning rate that is an unsupervised learning rule.

Abstract: A system and method relating to determining, by a processing device, a first number of parts waiting to be processed in a subsequent step by a plurality of machines housed in a plurality of buildings interconnected by rails, determining a second number of parts that the plurality of machines housed in the plurality of buildings is capable of processing in the subsequent step of the manufacture process over a determined period of time, calculating a capability occupancy ratio based on the first number and the second number for each one of the plurality of buildings, determining a target building of the plurality of buildings based on the capability occupancy ratio, and causing to dispatch the part to the target building via the rails.

Abstract: A convolutional neural network is provided comprising artificial neurons arranged in layers, each comprising output matrices. An output matrix comprises output neurons and is connected to an input matrix, comprising input neurons, by synapses associated with a convolution matrix comprising weight coefficients associated with the output neurons of an output matrix. Each synapse consists of a set of memristive devices storing a weight coefficient of the convolution matrix. In response to a change of the output value of an input neuron, the neural network dynamically associates each set of memristive devices with an output neuron connected to the input neuron. The neural network comprises accumulator(s) for each output neuron; to accumulate the values of the weight coefficients stored in the sets of memristive devices dynamically associated with the output neuron, the output value of the output neuron being determined from the value accumulated in the accumulator(s).

Abstract: An irrigation modeling framework in precision agriculture utilizes a combination of weather data, crop data, and other agricultural inputs to create customized agronomic models for diagnosing and predicting a moisture state in a field, and a corresponding need for, and timing of, irrigation activities. Specific combinations of various agricultural inputs can be applied, together with weather information to identify or adjust water-related characteristics of crops and soils, to model optimal irrigation activities and provide advisories, recommendations, and scheduling guidance for targeted application of artificial precipitation to address specific moisture conditions in a soil system of a field.

Abstract: An apparatus and method are provided to perform constrained optimization of a constrained property of an apparatus, which is complex due to having several components, and these components are configurable in real-time. The optimization is achieved by detecting values of the constrained property and a plurality of other properties of the apparatus when the apparatus is configured in a first subset of the plurality of configurations. A model is learned using the detected values of the constrained property. The model represents the constrained property and can also represent other properties as a function of the configurations. The model can also include estimated uncertainties of the constrained property in the model. Then, using the d model and the estimated uncertainties, the optimal configuration can be selected to minimize an error value (e.g., the difference between a desired value and an observed value of the at least one constrained property).

Abstract: A system and method of obtaining and utilizing an activity signature that is representative of a specific category of network activities based on directory service (DS) log data. The activity signature may be determining by a learning process, including segmenting and pruning a training dataset into a plurality of event segments and matching them with activities based on DS log data of known activities. Once obtained, the activity signature can advantageously be utilized to analyze any DS log data and activities in actual deployment. Using activity signatures to analyze DS event log can reveal roles of event-collection machines, aggregate information dispersed across their component events to reveal actors involved in particular AD activities, augment visibility of DS by enabling various vantage points to better infer activities at other domain machines, and reveal macro activities so that logged information becomes easily interpretable to human analysts.

Abstract: A building management system includes HVAC equipment operable to affect an indoor air temperature of a building, a system manager configured to obtain a cost function that characterizes a cost of operating the HVAC equipment, obtain a dataset relating to the building, determine a current state of the building by applying the dataset to a neural network, select a temperature bound associated with the current state, augment the cost function to include a penalty term that increases the cost when the indoor air temperature violates the temperature bound, and determine a temperature setpoint for each of a plurality of time steps in the future time period. The temperature setpoints achieve a target value of the cost function over the future time period. The building management system also includes a controller configured to operate the HVAC equipment to drive the indoor air temperature towards the temperature setpoint.

Abstract: A digital medium environment includes an action processing application that performs actions including personalized recommendation. A learning algorithm operates on a sample-by-sample basis (e.g., each instance a user visits a web page) and recommends an optimistic action, such as an action found by maximizing an expected reward, or a base action, such as an action from a baseline policy with known expected reward, subject to a safety constraint. The safety constraint requires that the expected performance of playing optimistic actions is at least as good as a predetermined percentage of the known performance of playing base actions. Thus, the learning algorithm is conservative during exploratory early stages of learning, and does not play unsafe actions. Furthermore, since the learning algorithm is online and can learn with each sample, it converges quickly and is able to track time varying parameters better than learning algorithms that learn on a block basis.

Abstract: Methods and systems are provided herein for calibrating subject data based on reference data, so that the calibrated subject data more closely represents a target population. The methods and systems include partitioning a reference data set into a plurality of reference data partitions using a data partitioning scheme, each reference data partition associated with a characteristic; and partitioning a subject data set into a plurality of subject data partitions using the data partitioning scheme, each subject data partition associated with a characteristic that corresponds to the characteristic associated with a reference data partition of the plurality of reference data partitions; identifying a variable present in the reference data set that is not present in the subject data set; and calculating a value of the variable for each reference data partition based on a rate of occurrence of the variable in each reference data partition.

Abstract: A method of machine learning includes performing dimensionality reduction on a parameter space by performing initial tests to determine scores for a plurality of parameter values in the parameter space, determining aggregate scores for a plurality of parameter value combinations, determining a ranking of the plurality of parameter value combinations based on the aggregate scores, and performing cluster analysis on the plurality of parameter value combinations to determine a set having highest aggregate scores. The method further includes performing additional tests, wherein each additional test is for a parameter value combination in the set. For each such parameter value combination, a probability of achieving a key performance indicator (KPI) is computed. Cluster analysis is then performed to determine a first subset of the set having highest probabilities of achieving the KPI. An operation is then performed on the first subset.