Abstract: An optimized computer architecture for training an neural network includes a system having multiple GPUs. The neural network may be divided into separate portions, and a different portion is assigned to each of the multiple GPUs. Within each GPU, its portion is further divided across multiple training worker threads in multiple processing cores, and each processing core has lock-free access to a local parameter memory. The local parameter memory of each GPU is separately, and individually, synchronized with a remote master parameter memory by lock memory access. Each GPU has a separate set of communication worker threads dedicated to data transfer between the GPU and the remote parameter memory so that the GPU's training worker threads are not involved with cross GPU communications.

Abstract: A content review system for an online system automatically determines if received content items to be displayed to users violate any policies of the online system. The content review system generates a semantic vector representing the semantic features of a content item, for example, using a neural network. By comparing the semantic vector for the content item with semantic vectors of content items previously determined to violate one or more policies, the content review system determines whether the content item also violates one or more policies. The content review system may also maintain templates corresponding to portions of semantic vectors shared by multiple content items. An analysis of historical content items that conform to the template is performed to determine a probability that received content items that conform to the template violate a policy.

Abstract: Systems, methods, and non-transitory computer readable media are configured to apply a machine learning model to predict an age division for a user based on user information. An age bracket within the age division including a largest number of connections of the user can be determined. The determined age bracket can be assigned for the user.

Abstract: A method is provided for rule creation that includes receiving (i) a MDP model with a set of states, a set of actions, and a set of transition probabilities, (ii) a policy that corresponds to rules for a rule engine, and (iii) a set of candidate states that can be added to the set of states. The method includes transforming the MDP model to include a reward function using an inverse reinforcement learning process on the MDP model and on the policy. The method includes finding a state from the candidate states, and generating a refined MDP model with the reward function by updating the transition probabilities related to the state. The method includes obtaining an optimal policy for the refined MDP model with the reward function, based on the reward policy, the state, and the updated probabilities. The method includes updating the rule engine based on the optimal policy.

Abstract: The present disclosure is drawn to the reduction of parameters in fully connected layers of neural networks. For a layer whose output is defined by y=Wx, where y?Rm is the output vector, x?Rn is the input vector, and W?Rm×n is a matrix of connection parameters, matrices Uij and Vij are defined and submatrices Wij are computed as the product of Uij and Vij, so that Wij=VijUij, and W is obtained by appending submatrices Wi,j.

Abstract: A spiking neural network (SNN) includes artificial neurons interconnected by artificial synapses to model a particular network. A first neuron emits spikes to neighboring neurons to cause a wave of spikes to propagate through the SNN. Weights of a portion of the synapses are increased responsive to the wave of spikes based on a spike timing dependent plasticity (STDP) rule. A second neuron emits spike to cause a chain of spikes to propagate to the first neuron on a path based on the increase in the synaptic weights. The path is determined to represent a shortest path in the particular network from a first network node represented by the second neuron to a second network node represented by the first neuron.

Abstract: A computer-implemented method for generating a simulated environment in which the behaviour of one or more individuals is modelled, the method comprising: defining a state of the environment at a first point in time; receiving an input defining an action to be performed by one or more individuals in the simulated environment; and in response to the input, updating the state of the environment based on a social-ecological model, wherein the social-ecological model simulates the behaviour of people within the environment and how the people respond to changes associated with said action, wherein the social-ecological model is a machine learning model that is trained using data reflective of real-life past events, the social-ecological model being configured to accept as input a parameterised dataset describing the state of the environment at the first point in time and to output an updated dataset that describes the updated state of the environment.

Abstract: Systems and methods for automated mathematical chatting. The systems and methods convert any identified non-numerical inputs into vectors and then perform the mathematical equation utilizing the vectors instead of the nonnumeric inputs along with any other identified numeric inputs to obtain a numerical vector result. The systems and methods decode the numerical vector result into a result feature and then search one or more databases for output based on the result feature. The systems and methods provide the selected output from the one or more databases in response to the mathematical query.

Abstract: Various systems and methods for implementing unsupervised or reinforcement learning operations for a neuron weight used in a neural network are described. In an example, the learning operations include processing a spike train input at a neuron of a spiking neural network, applying a synaptic weight, and observing spike events occurring before and after the neuron processing based on respective spike traces. A synaptic weight update process operates to generate a new value of the synaptic weight based upon the spike traces, configuration values, and a reference weight value. A reference weight update process also operates to generate a new value of the reference value for significant changes to the synaptic weight. Reinforcement may be provided in some examples to implement changes to the reference weight in reduced time. In some examples, the techniques may be implemented in a neuromorphic hardware implementation of the spiking neural network.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for providing actionable suggestions are disclosed. In one aspect, a method includes receiving (i) an indication that an event detection module has determined that a shared event of a particular type is presently occurring or has occurred, and (ii) data referencing an attribute associated with the shared event. The method includes selecting, from among multiple output templates that are each associated with a different type of shared event, a particular output template associated with the particular type of shared event detected by the module. The method generates a notification for output using at least (i) the selected particular output template, and (ii) the data referencing the attribute associated with the shared event. The method then provides, for output to a user device, the notification that is generated.

Abstract: A spike timing dependent plasticity (STDP) rule is applied in a spiking neural network (SNN) that includes artificial synapses bi-directionally connecting artificial neurons in the SNN to model locations within a physical environment. A first neuron is activated to cause a spike wave to propagate from the first neuron to other neurons in the SNN. Propagation of the spike wave causes synaptic weights of a subset of the synapses to be increased based on the STDP rule. A second neuron is activated after propagation of the spike wave to cause a spike chain to propagate along a path from the second neuron to the first neuron, based on the changes to the synaptic weights. A physical path is determined from the second to the first neuron based on the spike chain, and a signal may be sent to a controller of an autonomous device to cause the autonomous to navigate the physical path.

Abstract: The disclosed embodiments illustrate a domain adaptation method for learning transferable feature representations from a source domain for a target domain. The method includes receiving real-time input data comprising labeled instances of the source domain and unlabeled instances of the target domain from a computing device. The method further includes determining source specific representation corresponding to the source domain and a common representation shared between the source domain and the target domain. Based on a positive contribution from the source specific representation and the common representation, the labeled instances of the source domain are classified. The method further includes training a generalized classifier based on a positive contribution from the common representation. The method further includes automatically performing text classification on the unlabeled instances of the target domain based on the trained generalized classifier.

Abstract: Techniques to response to user requests using natural-language machine learning based on example conversations are described. In one embodiment, an apparatus may comprise a bot application interface component operative to receive an example-interaction repository, the example-interaction repository comprising a plurality of example user-to-bot interactions; and an interaction processing component operative to submit the example-interaction repository to a natural-language machine learning component; receive a sequence model from the natural-language machine learning component in response to submitting the example-interaction repository; and perform a user-to-bot conversation based on the sequence model. Other embodiments are described and claimed.

Abstract: A system and method for controlling a nodal network. The method includes estimating an effect on the objective caused by the existence or non-existence of a direct connection between a pair of nodes and changing a structure of the nodal network based at least in part on the estimate of the effect. A nodal network includes a strict partially ordered set, a weighted directed acyclic graph, an artificial neural network, and/or a layered feed-forward neural network.

Abstract: Systems and methods may include neuromorphic traffic control, such as between cores on a chip or between cores on different chips. The neuromorphic traffic control may include a plurality of routers organized in a mesh to transfer messages; and a plurality of neuron cores connected to the plurality of routers, the neuron cores in the plurality of neuron cores to advance in discrete time-steps, send spike messages to other neuron cores in the plurality of neuron cores during a time-step, and send barrier messages.

Abstract: A method for detecting road elements that may include (a) detecting predefined identifiers of road elements, in road related information sensed by vehicles; (b) detecting potential identifiers of road elements that differ from the predefined identifiers of road elements, by processing road related information that was acquired by the vehicles during relevant time windows that are related to the detecting of the predefined identifiers; (c) finding actual identifiers of road elements out of the potential identifiers; wherein the findings is based, at least in part, on road related information that was acquired by the vehicles outside the relevant time windows; and (d) updating a database with the actual identifiers.

Abstract: Techniques described herein include a method and system for item demand forecasting that utilizes machine learning techniques to generate a set of quantiles. In some embodiments, several item features may be identified as being relevant to an item forecast and may be provided as inputs to a regression module, which may calculate a set of quantiles for each item. A set of quantiles may comprise a number of confidence levels or probabilities associated with calculated demand values for an item. In some embodiments, costs associated with the item may be used to select an appropriate quantile associated (e.g., based on a corresponding confidence level). In some embodiments, an item demand forecast may be generated based on the calculated demand value associated with the selected quantile. In some embodiments, one or more of the item may be automatically ordered based on that item demand forecast.

Abstract: A method of training and using a machine learning model that controls for consideration of undesired factors which might otherwise be considered by the trained model during its subsequent analyses of new data. For example, the model may be a neural network trained on a set of training images to evaluate an insurance applicant based upon an image or audio data of the insurance applicant as part of an underwriting process to determine an appropriate life or health insurance premium. The model is trained to probabilistically correlate an aspect of the applicant's appearance with a personal and/or health-related characteristic. Any undesired factors, such as age, sex, ethnicity, and/or race, are identified for exclusion. The trained model receives the image (e.g., a “selfie”) of the insurance applicant, analyzes the image without considering the identified undesired factors, and suggests the appropriate insurance premium based only on the remaining desired factors.

Abstract: A robotic device is disclose as having deep reinforcement learning capability. The device includes non-transitory memory comprising instructions and one or more processors in communication with the memory. The instructions cause the one or more processors to receive a sensing frame, from a sensor, comprising an image. The processors then determine a movement transition based on the sensing frame and the deep reinforcement learning, wherein the deep reinforcement learning uses at least one of a map coverage reward, a map quality reward, or a traversability reward to determine the movement transition. The processors then update an area map based on the sensing frame and the deep reinforcement learning using a visual simultaneous localization and mapping (SLAM) process to determine the map updates.

Abstract: A system including: at least one processor; and at least one memory having stored thereon computer program code that, when executed by the at least one processor, controls the system to: receive a data model identification and a dataset; in response to determining that the data model does not contain a hierarchical structure, perform expectation propagation on the dataset to approximate the data model with a hierarchical structure; divide the dataset into a plurality of channels; for each of the plurality of channels: divide the data into a plurality of microbatches; process each microbatch of the plurality of microbatches through parallel iterators; and process the output of the parallel iterators through single-instruction multiple-data (SIMD) layers; and asynchronously merge results of the SIMD layers.