Abstract: The disclosed computer-implemented method may include (1) identifying an artificial neural network comprising a set of nodes interconnected via a set of connections, and (2) training the artificial neural network by, for each connection in the set of connections, determining a quantized weight value associated with the connection. Determining the quantized weight value associated with the connection may include (1) associating a loss function with the connection, the loss function including a periodic regularization function that describes a relationship between an input value and a weight value of the connection, (2) determining a minimum of the associated loss function with respect to the weight value in accordance with the periodic regularization function, and (3) generating the quantized weight value associated with the connection based on the determined minimum of the loss function. Various other methods, systems, and computer-readable media are also disclosed.

Abstract: A fuel system for an internal combustion engine includes a pressurized fuel supply, and a fuel injector fluidly connected to the pressurized fuel supply. The fuel injector includes a valve seat plate forming a valve seat, an injection control valve, and a nozzle cavity. A filter, which can include a 2-dimensional filter, is positioned partially within the valve seat plate and partially within an injector body to filter an incoming flow of high-pressure fuel. The valve seat plate and filter are integrated in a valve seat plate and filter assembly.

Abstract: A smoking device has an inner cylinder receivable within an outer cylinder, the inner cylinder having a combustion chamber, an air chamber, a mouthpiece, and one or more air chamber slots and combustion chamber slots; the outer cylinder having a bore hole for receiving the inner cylinder, a smoke channel, and one or more bore slots. In some examples, the inner cylinder and outer cylinder have one or more magnets for securing the inner cylinder to the outer cylinder and for aligning the air and combustion chamber slots with the bore slots. A user may rotate the inner cylinder in relation to the outer cylinder, thereby changing which air and combustion slots are aligned with the bore slots and the smoke channel. As a result of the various alignments, a user is able to vary the temperature, taste, fuel-to-air ratio, and other factors.

Abstract: A treatment mechanism for dispensing treatment fluid onto one or more plants in a field is described. The treatment mechanism includes a valve for regulating the dispensing of treatment fluid and a nozzle holder comprising a plurality of nozzles. Each nozzle is configured to dispense treatment fluid and to couple to the valve. The treatment mechanism further includes a control system configured to receive a plant treatment instruction for treating the plants. The plant treatment instruction includes a treatment position for the nozzle holder. The control system is further configured to determine a current position of the nozzle holder and adjust the current position of the nozzle holder to the treatment position for the nozzle holder. The control system is further configured to actuate the treatment mechanism such that the plants are treated via a nozzle of the plurality of nozzles.

Abstract: Systems and methods are described for implementing an “arc jump” technique in conjunction with bounded loads in consistent hashing. In general, bounded loads refers to limiting the ability of a single device within a distributed system to store data objects, such that when a request to store a new data object would otherwise be directed to that device, it is instead redirected to an alternative device. Redirecting all requests to a single alternative device can lead to cascading failures, as the alternative device must maintain its own load and that which has been redirected to it. Embodiments of the present disclosure address this by determining an alternative device on a per-object basis, such as by again hashing the object with an additional seed value. This distributes request from an overloaded device among all other devices of the distributed system, avoiding cascading failures.

Abstract: Systems and methods are described for load balancing requests in a distributed system using consistent hashing. Specifically, systems and methods are described for using “the power of k choices” when placing new servers on a consistent hash ring used to load balance requests. Rather than placing each new server at a fixed point determined by a hashing algorithm, a load balancer can identify multiple potential points on the hash ring for the new server. The load balancer can then compare these points to determine a preferred location, and place the server at the preferred location. Techniques described herein can substantially improve placement of servers, which in turn results in better load balancing.

Abstract: Techniques for comparing customer records to identify linked customer records are provided. The techniques may include a computer system identifying a pair of customer records and comparing the customer records to determine if the customer records are linked records. The computer system may analyze a first set of data values from first corresponding data fields from the pair of records to determine a first distance between the first set of data values, analyze a second set of data values from second corresponding data fields from the pair of records to determine a second distance between the second set of data values, and combine the first distance and the second distance into a combined distance value which is representative of a distance between the pair of customer records by using a combination function which provides a maximum value cap for the first distance and a weight for the first distance.

Abstract: In one embodiment, a method includes, by a computing device, receiving a plurality of inputs for a convolution layer of a convolutional neural network, the convolution layer having one or more input channels and one or more output channels, wherein the inputs are received via the input channels, generating, by convolving the inputs with one or more two-dimensional filters, a plurality of intermediate values, and generating, by convolving the intermediate values with one or more one-dimensional filters, a plurality of outputs, wherein the one-dimensional filters receive the intermediate values from the two-dimensional filters via intermediate channels. The method may provide the outputs to a subsequent layer of the convolutional neural network via the output channels. Each of the two dimensions of the two-dimensional filter may correspond to a spatial dimension, and the one dimension of the one-dimensional filter may correspond to a temporal dimension.

Abstract: Systems, methods, and non-transitory computer-readable media can acquire an image that depicts at least one character. A set of pixels, within the image, through which the at least one character is depicted can be identified. At least one linear portion, within the image, can be identified based on the set of pixels. For each sub-portion within the at least one linear portion, a respective first confidence score representing a respective first likelihood that a respective sub-portion depicts the at least one character can be determined.

Abstract: Embodiments of a system as described herein may receive financial institution product information from a plurality of financial institutions distributed across a computing network. The system may also receive data from the plurality of financial institutions distributed across the computer network and create or update an ontology. A relevance score may be generated for a set of financial institution products which may, in conjunction with a campaign definition provided by a financial institution administrator, be used to associate users with a list of campaigns which may be stored as campaign data. An online banking application at a user device may request campaign data for a user. In response, the system may return campaign data for the user to the online banking application. Using the campaign data, the online banking application may select one or more products to recommend to the user and display content for the selected products on the user device.

Abstract: In one embodiment, a method includes a computing system accessing a first training data comprising a first image and a second image and an associated optical flow estimation. The system may input (1) the first image into a first machine-learning model configured to generate a first output and (2) the optical flow estimation into a second machine-learning model configured to generate a second output. The first output of the first machine-learning model is associated with first image segments of a predetermined number, and the second output of the second machine-learning model is associated with transformations of the predetermined number. The first output, the transformations, and the first image are configured to generate an estimated image. The system trains the first machine-learning model and the second machine-learning model based on at least a comparison of the estimated image and the second image.

Abstract: Embodiments of a system as described herein may receive financial institution product information from a plurality of financial institutions distributed across a computing network. The system may also receive data from the plurality of financial institutions distributed across the computer network and create or update an ontology. A relevance score may be generated for a set of financial institution products which may, in conjunction with a campaign definition provided by a financial institution administrator, be used to associate users with a list of campaigns which may be stored as campaign data. An online banking application at a user device may request campaign data for a user. In response, the system may return campaign data for the user to the online banking application. Using the campaign data, the online banking application may select one or more products to recommend to the user and display content for the selected products on the user device.

Abstract: In one embodiment, a method includes a computing system accessing a first training data comprising a first image and a second image and an associated optical flow estimation. The system may input (1) the first image into a first machine-learning model configured to generate a first output and (2) the optical flow estimation into a second machine-learning model configured to generate a second output. The first output of the first machine-learning model is associated with first image segments of a predetermined number, and the second output of the second machine-learning model is associated with transformations of the predetermined number. The first output, the transformations, and the first image are configured to generate an estimated image. The system trains the first machine-learning model and the second machine-learning model based on at least a comparison of the estimated image and the second image.

Abstract: Systems, methods, and non-transitory computer-readable media can identify a media content item for which media processing is to be performed. State information associated with the media content item can be acquired. At least some of the media processing can be enabled, based on the state information, to be performed client-side with respect to the media content item. The state information can indicate a next processing step of the at least some of the media processing that is to be performed. The state information can be updated based on the at least some of the media processing performed client-side.

Abstract: The present disclosure extends to temporally consistent group keys for groups of linked records in a data store. In embodiments, records in the data store can be categorized as persistent or temporary. Each record carries a unique record identifier and a timestamp depicting the record's time of creation. The group key for a group is the record identifier for the oldest persistent record in the group. If a group has no persistent records, the group key is the record identifier for the oldest temporary record in the group. If two or more records have identical timestamps, the group key is the first record identifier in a lexicographical ordering of the record identifiers for records having the earliest timestamps.

Abstract: Systems, methods, and non-transitory computer-readable media can acquire an image that depicts at least one character. A set of pixels, within the image, through which the at least one character is depicted can be identified. At least one linear portion, within the image, can be identified based on the set of pixels. For each sub-portion within the at least one linear portion, a respective first confidence score representing a respective first likelihood that a respective sub-portion depicts the at least one character can be determined.

Abstract: Systems, methods, and non-transitory computer-readable media can identify a media content item for which media processing is to be performed. State information associated with the media content item can be acquired. At least some of the media processing can be enabled, based on the state information, to be performed client-side with respect to the media content item. The state information can indicate a next processing step of the at least some of the media processing that is to be performed. The state information can be updated based on the at least some of the media processing performed client-side.

Abstract: Techniques for comparing customer records to identify linked customer records are provided. The techniques may include a computer system identifying a target group of electronic customer records which have data pertaining to a customer, receiving a hierarchical blocker, and analyzing a customer record against the hierarchical blocker and extracting one or more data values from data fields identified by the hierarchical blocker. The computer system may create block keys from the extracted data values, associate the block keys with the customer record, create a block key group of customer records which are associated with a particular block key, and compare pairs of records from the block key group to identify records which are likely to be linked records because they pertain to a single customer entity.

Abstract: Techniques for comparing customer records to identify linked customer records pertaining to a single customer entity are provided. The techniques include identifying a target group of electronic customer records having data fields containing data pertaining to a customer, identifying one or more focused blockers identifying a data value for an electronic customer record data field, and analyzing the target group of electronic customer records to identify a focused group of electronic customer records containing the focused blocker data value. The techniques also include comparing pairs of electronic customer records from the focused group of electronic customer records to identify linked records which pertain to a single customer entity.

Abstract: The present disclosure extends to comparing two or more names in a database of contact records. In embodiments, systems of the present disclosure execute a comparison of submitted names to determine if the records should be linked and/or merged. Embodiments of the present disclosure can compensate for irregularities in the database of contact records, including inconsistent name prefixes and suffixes, misspellings, typographical errors, misordered first/last names, and nicknames.