Abstract: Systems, methods, and computer program products for performing semi-supervised contrastive learning of visual representations are provided. For example, the present disclosure provides systems and methods that leverage particular data augmentation schemes and a learnable nonlinear transformation between the representation and the contrastive loss to provide improved visual representations. Further, the present disclosure also provides improvements for semi-supervised contrastive learning.

Abstract: Systems, methods, and computer program products for performing semi-supervised contrastive learning of visual representations are provided. For example, the present disclosure provides systems and methods that leverage particular data augmentation schemes and a learnable nonlinear transformation between the representation and the contrastive loss to provide improved visual representations. Further, the present disclosure also provides improvements for semi-supervised contrastive learning.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, relating to multi-task recurrent neural networks. One of the methods includes maintaining data specifying, for a recurrent neural network, a separate internal state for each of a plurality of memory regions; receiving a current input; identifying a particular memory region of the memory access address defined by the current input; selecting, from the internal states specified in the maintained data, the internal state for the particular memory region; processing, in accordance with the selected internal state for the particular memory region, the current input in the sequence of inputs using the recurrent neural network to: generate an output, the output defining a probability distribution of a predicted memory access address, and update the selected internal state of the particular memory region; and associating the updated selected internal state with the particular memory region in the maintained data.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for pre-fetching data from memory using neural networks. One example system receives a sequence of prior program counter addresses of a computer program and corresponding delta values. The system creates an input representation based on the sequence. The system provides the input representation as input to a recurrent neural network. The system receives from the recurrent neural network an output that defines a probability distribution over future delta values. Each probability in the distribution represents a likelihood that execution of a future instruction of the computer program will cause data to be fetched from a particular future memory address.

Abstract: Methods and apparatus are disclosed for a coupling nozzle for cryogenic fluid. A nozzle for cryogenic fluid includes a flow body, a flow control assembly, a first linear actuator, and a locking actuator. The flow body defines a conduit. The flow control assembly is at least partially disposed in the conduit of the flow body. The flow control assembly is configured to transition between an open position and a closed position to permit and prevent, respectively, the cryogenic fluid to flow through the flow body. The first linear actuator includes an actuator body and a first shaft. The first shaft is configured to slide between an extended position and a contracted position to transition the flow control assembly between the open position and the closed position, respectively. The locking actuator is for locking the nozzle to a receptacle in an automated manner.

Abstract: Methods and apparatus are disclosed for a coupling nozzle for cryogenic fluid. A coupling nozzle includes a flow body. The coupling nozzle includes a mount, a flow control assembly, and a pneumatic cylinder. The pneumatic cylinder includes a cylinder body and a shaft. The shaft is coupled to and configured to actuate the flow body. The coupling nozzle includes a first locking mechanism coupled to the mount and configured to secure the coupling nozzle to a receptacle in a locked position. When the first locking mechanism is in the locked position, the shaft is configured to actuate the flow control assembly. The coupling nozzle includes a redundant locking mechanism including one or more feet and a lock. The one or more feet are configured to engage the lock to prevent the first locking mechanism from transitioning from the locked position when the shaft is in the extended position.

Abstract: A low-emission nozzle and receptacle coupling for cryogenic fluid is disclosed. A nozzle includes a body, a shaft, a poppet eat, and a poppet. The body includes an outer shell, body segment(s), and interior walls sealingly coupled together. At least one of the interior walls is coupled to the outer shell. The interior walls extend longitudinally back-and-forth in a zig-zag pattern to define an elongated conduction path between the chamber and an exterior of the nozzle to impede a heat leak between the chamber and the exterior. The shaft is housed within and slidably extends through the chamber. The poppet is coupled to the shaft and is configured to engage a receptacle poppet when the receptacle is coupled to the nozzle. The poppet is configured to engage the poppet seat in a closed position and be disengaged from the poppet seat in an open position.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, relating to multi-task recurrent neural networks. One of the methods includes maintaining data specifying, for a recurrent neural network, a separate internal state for each of a plurality of memory regions; receiving a current input; identifying a particular memory region of the memory access address defined by the current input; selecting, from the internal states specified in the maintained data, the internal state for the particular memory region; processing, in accordance with the selected internal state for the particular memory region, the current input in the sequence of inputs using the recurrent neural network to: generate an output, the output defining a probability distribution of a predicted memory access address, and update the selected internal state of the particular memory region; and associating the updated selected internal state with the particular memory region in the maintained data.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for processing an input image using a super-resolution neural network to generate an up-sampled image that is a higher resolution version of the input image. In one aspect, a method comprises: processing the input image using an encoder subnetwork of the super-resolution neural network to generate a feature map; generating an updated feature map, comprising, for each spatial position in the updated feature map: applying a convolutional filter to the feature map to generate a plurality of features corresponding to the spatial position in the updated feature map, wherein the convolutional filter is parametrized by a set of convolutional filter parameters that are generated by processing data representing the spatial position using a hyper neural network; and processing the updated feature map using a projection subnetwork of the super-resolution neural network to generate the up-sampled image.

Abstract: A low-emission nozzle and receptacle coupling for cryogenic fluid is disclosed. An example nozzle includes a body including a front end and a back end and defining a chamber through which the cryogenic fluid is to flow to the receptacle. The nozzle includes a shaft having a first end and a second end. The shaft is housed within and slidably extending through the chamber. The nozzle includes a poppet coupled to the first end of the shaft and an actuator including a stem coupled to the second end of the shaft. The stem is configured to linearly actuate to cause the shaft and the poppet to linearly actuate. The nozzle includes a coupling assembly coupling the actuator to the back end of the body. The coupling assembly includes insulating material to thermally isolate the actuator from the chamber through which the cryogenic fluid is to flow.

Abstract: A cleaning receptacle for cleaning a fueling nozzle used to transfer cryogenic fluid is disclosed. An example cleaning receptacle includes an outer body. The outer body includes an outer wall that extends circumferentially around a center axis of the cleaning receptacle. The outer wall has a first inner surface. The outer body includes a plurality of locating members extending inwardly from the first inner surface toward the center axis. The example cleaning receptacle includes a flow body. The flow body includes a second inner surface that defines a cavity and a lip at an outer end of the flow body. The example cleaning receptacle includes a spray body disposed at an inner end of the flow body. The spray body includes a spray head that defines a plurality of spray holes configured to spray air through the cavity of the flow body and onto surfaces of the fueling nozzle.

Abstract: A cleaning nozzle for cleaning a fueling receptacle capable of receiving a separate fueling nozzle for transferring cryogenic fluid from a fueling station is disclosed. An example cleaning nozzle includes a body that defines a cavity. The example cleaning nozzle includes a nozzle head coupled to and extending from a distal end of the body. The nozzle head defines a plurality of spray holes to spray pressurized air onto the fueling receptacle. The example cleaning nozzle includes a blower that defines a blower inlet, a blower outlet, and a blower flow path extending between the blower inlet and the blower outlet. The blower comprises a valve and a lever operatively coupled to the valve. The example cleaning nozzle includes an insert defining an insert flow path that is fluidly coupled to the blower outlet. The example cleaning nozzle includes a connector fluidly connecting the insert and the nozzle head.

Abstract: Methods and apparatus are disclosed for a low-emission nozzle and receptacle coupling for cryogenic fluid. An example nozzle includes a body defining a chamber through which cryogenic fluid is to flow. The body includes an outer shell that includes an outer shell surface. The nozzle includes a locking assembly configured to securely couple the nozzle to a receptacle. The locking assembly includes an inner sleeve fixedly coupled to the outer shell surface and an outer sleeve extending over and rotatably coupled to the inner sleeve. One or more locking teeth are fixedly coupled to the outer sleeve and configured to be slidably received by respective one or more coupling slots of the receptacle. The one or more locking teeth are configured to rotatably slide within the respective one or more coupling slots to couple the nozzle to the receptacle as the outer sleeve rotates relative to the inner sleeve.

Abstract: Systems, methods, and computer program products for performing semi-supervised contrastive learning of visual representations are provided. For example, the present disclosure provides systems and methods that leverage particular data augmentation schemes and a learnable nonlinear transformation between the representation and the contrastive loss to provide improved visual representations. Further, the present disclosure also provides improvements for semi-supervised contrastive learning.

Abstract: Methods and apparatus are disclosed for a coupling nozzle for cryogenic fluid. A coupling nozzle includes a flow body. The coupling nozzle includes a mount, a flow control assembly, and a pneumatic cylinder. The pneumatic cylinder includes a cylinder body and a shaft. The shaft is coupled to and configured to actuate the flow body. The coupling nozzle includes a first locking mechanism coupled to the mount and configured to secure the coupling nozzle to a receptacle in a locked position. When the first locking mechanism is in the locked position, the shaft is configured to actuate the flow control assembly. The coupling nozzle includes a redundant locking mechanism including one or more feet and a lock. The one or more feet are configured to engage the lock to prevent the first locking mechanism from transitioning from the locked position when the shaft is in the extended position.

Abstract: Methods and apparatus are disclosed for a coupling nozzle for cryogenic fluid. An example nozzle includes a flow body defining a conduit. The flow body is configured to permit cryogenic fluid to flow through the conduit. The nozzle includes a mounting ring through which the flow body slidably extends and a bushing fixedly positioned adjacent the mounting ring. The bushing slidably receives the flow body in a keyed manner to prevent rotation of the flow body. The nozzle includes a flow control assembly at least partially disposed in the conduit of the flow body. The flow control assembly is configured to permit the cryogenic fluid to flow through the flow body in an open position and prevent the cryogenic fluid from flowing through the flow body in a closed position.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, relating to multi-task recurrent neural networks. One of the methods includes maintaining data specifying, for a recurrent neural network, a separate internal state for each of a plurality of memory regions; receiving a current input; identifying a particular memory region of the memory access address defined by the current input; selecting, from the internal states specified in the maintained data, the internal state for the particular memory region; processing, in accordance with the selected internal state for the particular memory region, the current input in the sequence of inputs using the recurrent neural network to: generate an output, the output defining a probability distribution of a predicted memory access address, and update the selected internal state of the particular memory region; and associating the updated selected internal state with the particular memory region in the maintained data.

Abstract: Methods and apparatus are disclosed for a low-emission nozzle and receptacle coupling for cryogenic fluid. An example nozzle includes a body defining a chamber through which cryogenic fluid is to flow. The body includes an outer shell that includes an outer shell surface. The nozzle includes a locking assembly configured to securely couple the nozzle to a receptacle. The locking assembly includes an inner sleeve fixedly coupled to the outer shell surface and an outer sleeve extending over and rotatably coupled to the inner sleeve. One or more locking teeth are fixedly coupled to the outer sleeve and configured to be slidably received by respective one or more coupling slots of the receptacle. The one or more locking teeth are configured to rotatably slide within the respective one or more coupling slots to couple the nozzle to the receptacle as the outer sleeve rotates relative to the inner sleeve.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, relating to multi-task recurrent neural networks. One of the methods includes maintaining data specifying, for a recurrent neural network, a separate internal state for each of a plurality of memory regions; receiving a current input; identifying a particular memory region of the memory access address defined by the current input; selecting, from the internal states specified in the maintained data, the internal state for the particular memory region; processing, in accordance with the selected internal state for the particular memory region, the current input in the sequence of inputs using the recurrent neural network to: generate an output, the output defining a probability distribution of a predicted memory access address, and update the selected internal state of the particular memory region; and associating the updated selected internal state with the particular memory region in the maintained data.