Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for a neural network system comprising one or more gated linear networks. A system includes: one or more gated linear networks, wherein each gated linear network corresponds to a respective data value in an output data sample and is configured to generate a network probability output that defines a probability distribution over possible values for the corresponding data value, wherein each gated linear network comprises a plurality of layers, wherein the plurality of layers comprises a plurality of gated linear layers, wherein each gated linear layer has one or more nodes, and wherein each node is configured to: receive a plurality of inputs, receive side information for the node; combine the plurality of inputs according to a set of weights defined by the side information, and generate and output a node probability output for the corresponding data value.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for a neural network system comprising one or more gated linear networks. A system includes: one or more gated linear networks, wherein each gated linear network corresponds to a respective data value in an output data sample and is configured to generate a network probability output that defines a probability distribution over possible values for the corresponding data value, wherein each gated linear network comprises a plurality of layers, wherein the plurality of layers comprises a plurality of gated linear layers, wherein each gated linear layer has one or more nodes, and wherein each node is configured to: receive a plurality of inputs, receive side information for the node; combine the plurality of inputs according to a set of weights defined by the side information, and generate and output a node probability output for the corresponding data value.

Abstract: A netflow generator appliance is provided. The netflow generator appliance is configured to perform operations, including receiving a plurality of netflow records at a netflow generator appliance, each netflow record including at least a hash tag; converting a first hash tag of the first netflow record into a first percentage based on a load balancing function; based on the first percentage, storing the first netflow record in a first queue of data that are scheduled to be sent to a first netflow collector; converting a second hash tag of a second netflow record into a second percentage based on the load balancing function, wherein the second percentage differs from the first percentage; and based on the second percentage, storing the second netflow record in a second queue of data that are scheduled to be sent to a second netflow collector.

Abstract: A computer-implemented method of distributing netflow records is disclosed. The method includes receiving a plurality of netflow records, each netflow record associated with a hash tag. The hash tag may include, for example, a random M-bit pattern of bits. The method filters the plurality of netflow records over a first hash tag range in response to the hash tag value and storing the first filtered netflow records in a first queue associated with a first netflow collector; filters the plurality of netflow records over a second hash tag range in response to the hash tag value and storing the second filtered netflow records in a second queue associated with a second netflow collector, wherein the second hash tag range differs from the first hash tag range; sends the first filtered netflow records from the first queue to the first netflow collector; and sends the second filtered netflow records from the second queue to the second netflow collector.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for a neural network system comprising one or more gated linear networks. A system includes: one or more gated linear networks, wherein each gated linear network corresponds to a respective data value in an output data sample and is configured to generate a network probability output that defines a probability distribution over possible values for the corresponding data value, wherein each gated linear network comprises a plurality of layers, wherein the plurality of layers comprises a plurality of gated linear layers, wherein each gated linear layer has one or more nodes, and wherein each node is configured to: receive a plurality of inputs, receive side information for the node; combine the plurality of inputs according to a set of weights defined by the side information, and generate and output a node probability output for the corresponding data value.

Abstract: A computer-implemented method of distributing netflow records is disclosed. The method includes receiving a plurality of netflow records, each netflow record associated with a hash tag. The hash tag may include, for example, a random M-bit pattern of bits. The method filters the plurality of netflow records over a first hash tag range in response to the hash tag value and storing the first filtered netflow records in a first queue associated with a first netflow collector; filters the plurality of netflow records over a second hash tag range in response to the hash tag value and storing the second filtered netflow records in a second queue associated with a second netflow collector, wherein the second hash tag range differs from the first hash tag range; sends the first filtered netflow records from the first queue to the first netflow collector; and sends the second filtered netflow records from the second queue to the second netflow collector.

Abstract: A computer-implemented method of distributing netflow records is disclosed. The method includes receiving a plurality of netflow records, each netflow record associated with a hash tag. The hash tag may include, for example, a random M-bit pattern of bits. The method filters the plurality of netflow records over a first hash tag range in response to the hash tag value and storing the first filtered netflow records in a first queue associated with a first netflow collector; filters the plurality of netflow records over a second hash tag range in response to the hash tag value and storing the second filtered netflow records in a second queue associated with a second netflow collector, wherein the second hash tag range differs from the first hash tag range; sends the first filtered netflow records from the first queue to the first netflow collector; and sends the second filtered netflow records from the second queue to the second netflow collector.

Abstract: A computer-implemented method of distributing netflow records is disclosed. The method includes receiving a plurality of netflow records, each netflow record associated with a hash tag. The hash tag may include, for example, a random M-bit pattern of bits. The method filters the plurality of netflow records over a first hash tag range in response to the hash tag value and storing the first filtered netflow records in a first queue associated with a first netflow collector; filters the plurality of netflow records over a second hash tag range in response to the hash tag value and storing the second filtered netflow records in a second queue associated with a second netflow collector, wherein the second hash tag range differs from the first hash tag range; sends the first filtered netflow records from the first queue to the first netflow collector; and sends the second filtered netflow records from the second queue to the second netflow collector.

Abstract: A computer-implemented method of distributing netflow records is disclosed. The method includes receiving a plurality of netflow records, each netflow record associated with a hash tag. The hash tag may include, for example, a random M-bit pattern of bits. The method filters the plurality of netflow records over a first hash tag range in response to the hash tag value and storing the first filtered netflow records in a first queue associated with a first netflow collector; filters the plurality of netflow records over a second hash tag range in response to the hash tag value and storing the second filtered netflow records in a second queue associated with a second netflow collector, wherein the second hash tag range differs from the first hash tag range; sends the first filtered netflow records from the first queue to the first netflow collector; and sends the second filtered netflow records from the second queue to the second netflow collector.

Abstract: The invention concerns a method for compressing a digitally coded video frame sequence. In the method, a given frame is divided into blocks, and the information content of selected blocks is modified, relying on information contained in a neighboring block or blocks (prediction), and the blocks are converted from spatial representation into frequency representation. The information content of the transformed blocks is encoded by arithmetic coding. The efficiency of the coding is improved by various methods, such as dynamically partitioning the blocks into sub-blocks, or performing a compressibility analysis is the blocks before carrying out further transformations. The entropy coding uses a neural network to determine the parameters of the arithmetic coding. The frames are dynamically re-scaled, depending on available bandwidth and quality of the coded image.

Abstract: There is disclosed a method and apparatus for coding data, for the purpose of preventing unauthorised readout of information. In the method, user information is converted into digital data (UD), and the digital data is recorded in a first data set (CD) and a second data set (HD), where the first data set (CD) is recorded in a first optical storage medium of a data carrier, and the second data set (HD), is recorded in a second, holographic storage medium of the data carrier. According to the method, the digital data (UD) is divided between the first and second data sets (CD,HD) in a manner so that a part of the user information is contained only in the first data set (CD), and another part of the user information is contained only in the second data set (HD), and the relative proportion of the user information in the data sets is selected so that the information contained in any one of the data sets is insufficient by itself to restore the complete user information (UD) contained in both data sets.

Abstract: The invention concerns a method for compressing a digitally coded video frame sequence. In the method, a given frame is divided into blocks, and the information content of selected blocks is modified, relying on information contained in a neighbouring block or blocks (prediction), and the blocks are converted from spatial representation into frequency representation. The information content of the transformed blocks is encoded by arithmetic coding. The efficiency of the coding is improved by various methods, such as dynamically partitioning the blocks into sub-blocks, or performing a compressibility analysis is the blocks before carrying out further transformations. The entropy coding uses a neural network to determine the parameters of the arithmetic coding. The frames are dynamically re-scaled, depending on available bandwidth and quality of the coded image.

Abstract: A tank heater control that will control a heater used for heating pressurised tanks over a broad range of temperatures. The system uses a sensor for providing a signal to a processor, in which the signal is indicative of flow of gas from the tank. The processor energizes the heater when gas flows from the tank. In various embodiments, the sensor may be a flow sensor on an outlet from the tank, temperature sensors that detect a difference in temperature of the tank from ambient temperature, or a pressure sensor on an outlet from the tank.

Abstract: There is disclosed a method and apparatus for coding data, for the purpose of preventing unauthorised readout of information. In the method, user information is converted into digital data (UD), and the digital data is recorded in a first data set (CD) and a second data set (HD), where the first data set (CD) is recorded in a first optical storage medium of a data carrier, and the second data set (HD), is recorded in a second, holographic storage medium of the data carrier. According to the method, the digital data (UD) is divided between the first and second data sets (CD,HD) in a manner so that a part of the user information is contained only in the first data set (CD), and another part of the user information is contained only in the second data set (HD), and the relative proportion of the user information in the data sets is selected so that the information contained in any one of the data sets is insufficient by itself to restore the complete user information (UD) contained in both data sets.

Abstract: There is disclosed a method for the holographic recording of data, wherein a hologram containing the date is recorded in a waveguide layer (3). The holograms are formed in a layer structure containing multiple waveguide layers (3). Coupling means (12) are provided for selectively coupling the reference beam (11) into the single waveguide layers (3) of the layer structure. The invention also relates to an arrangement for holographic recording of data, comprising a data storage medium with a waveguide holographic storage layer, and an optical system for writing and reading the holograms. The optical system comprises means for imaging an object beam (5) and a reference beam (11) on the storage medium. The arrangement comprises multiple waveguide holographic storage layers (3) in the storage medium, and means for selectively coupling the reference beam (11) into the single waveguide layers (3) of the layer structure.

Abstract: There is disclosed a method for the holographic recording of data, wherein a hologram containing the date is recorded in a waveguide layer (3). The holograms are formed in a layer structure containing multiple waveguide layers (3). Coupling means (12) are provided for selectively coupling the reference beam (11) into the single waveguide layers (3) of the layer structure. The invention also relates to an arrangement for holographic recording of data, comprising a data storage medium with a waveguide holographic storage layer, and an optical system for writing and reading the holograms. The optical system comprises means for imaging an object beam (5) and a reference beam (11) on the storage medium. The arrangement comprises multiple waveguide holographic storage layers (3) in the storage medium, and means for selectively coupling the reference beam (11) into the single waveguide layers (3) of the layer structure.

Abstract: The method is employed for correcting aiming errors between devices of fire control systems and weapons installations by aiming guns (G) with target measuring sensors (Sg) on a common measuring target (K, Ki) by means of a target tracking device (T) in order to detect deviation values (Di) between the position of the measuring target (Ki) and of the gun (G) controlled by the target tracking device (T), which is represented by the aiming (O) of the target measuring sensor (Sg). Following the evaluation of the position deviation, an aiming error vector (B) is processed for being taken into account in a servo gun control. The correction of the aiming error vector (B) is performed recursively in accordance with the method of the least error squares.

Abstract: Method of storing and retrieving information by exposing cells of a storage medium to a first beam of light having a first set of properties affecting the optical properties of said storage medium, and by exposing a plurality of individual cells, forming a page of information, of said storage medium to a second beam of light having a second set of properties. A third light beam from said storage medium forms an image of said page on a detecting means. The detected light image is converted into electric data signals, each bit of said data signals corresponding to a cell of said storage medium. The first beam of light is generated by a first light source and there is provided an electrooptical means having an electric input for receiving data and being arranged for modulating said first light beam in dependance of said data.

Abstract: The invention relates to a read-write optical memory comprising a plurality of juxtaposed memory cells (11), each receiving a respective light beam (3). Each memory cell contains a storage medium (10), which includes a storage element (23) having stable optical states. The storage element (23) is divided into a number of memory points, and the optical state in a given memory point can be both changed and read by means of a light beam (3) directed towards the memory point. The memory can be implemented entirely without any movable mechanical parts and has a very short read-write time and an exceptionally high storage capacity. Parallel writing and reading of multibit words is possible.

Abstract: A weapons system with at least one fire control device and a launcher uses a remotely fragmentable projectile for defense against a missile or other movable target. Although the probability of a hit is increased with the use of a remotely fragmentable projectile, the system of the invention increases the success of the defense, which is nevertheless questionable because of the low density of the projectile fragments. The method utilizes a projectile, the fragments of which are concentrated in a ring, which spreads out in the shape of a conical envelope. The projectile is individualized at launch. Tracking of the target continues during the flight of the projectile, because of which the expected location at the pre-calculated time of impact becomes continuously better known. The fragmentation command is transmitted to the projectile as late as possible. It is always possible with a high degree of probability to score a hit.