Abstract: A racially unbiased mammogram analyzer includes an interface for receiving mammograms; a processor for extracting features of mammograms of general population; a processor for extracting features of mammograms of a specific race. In one embodiment, the general population mammogram features are represented by middle layers of a CNN and the race specific features are represented by the end layer of the CNN network. In one embodiment, the race specific layers of CNN change dynamically according to the race indication done explicitly. In one embodiment the race specific layers of CNN change dynamically according to the race indication given by race indication processor. In one embodiment, the race indications are computed by a network of parallel variational autoencoder networks. In one embodiment, the race indicator computes race specific information to the CNN and are provided by variational autoencoders.

Abstract: The invention includes systems and methods for improving the performance of non-blocking data switching systems. One embodiment of the invention includes a method comprising routing data from a plurality of inputs to a plurality of outputs through a switching core according to a first switching schedule, receiving a first set of reports comprising reports from data sources associated with the plurality of inputs, evaluating one or more reports of the first set of reports, determining a sufficiency of the first switching schedule based on the one or more reports, adapting a second switching schedule, wherein the second switching schedule differs from the first switching schedule, sending the second switching schedule to the data sources, issuing one or more synchronization signals associated with a transition to the second switching schedule to the data sources and routing data from the plurality of inputs to the plurality of outputs through the switching core according to the second switching schedule.

Abstract: Methods and systems for dynamically computing a schedule defining interconnections between ports in a switching system. In one embodiment, a switching core requests demand reports from the ingress ports. In response to this request, the ingress ports generate a series of suggested port schedules, beginning with a first choice, then a second choice, and so on. The schedules are transmitted to the switching core, beginning with the first choice. The switching core receives the first choice for each of the ports and determines whether the aggregate schedule defined by the first choices is valid. If the aggregate schedule is valid, then this schedule is implemented. If the aggregate schedule is not valid, portions of it are discarded and the next choice from each of the ports is examined to identify connections to replace the discarded portions. This process is repeated until a valid schedule is obtained.

Abstract: The present invention provides a system and method for slot deflection routing of optical data packets. The method of the present invention includes the steps of establishing a schedule pattern that includes a plurality of time slots. The schedule pattern includes at least one time slot in which an ingress edge unit can communicate with a destination egress edge unit, at least one time slot in which the ingress edge unit can communicate with a intermediate edge unit, and at least one time slot in which the intermediate edge unit can communicate with the destination egress edge unit. The present invention also includes receiving a data packet at the ingress edge unit and determining if the schedule pattern allocates sufficient bandwidth to send the data packet from the ingress edge unit to the destination egress edge unit without deflecting the data packet through an intermediate edge unit.

Abstract: A non-blocking optical matrix core switching method that includes maintaining a schedule for routing data through an optical matrix core and receiving and analyzing reports from peripheral devices. The method determines whether the schedule is adequate for the current data traffic patterns and if the schedule is not adequate a new schedule is implemented. The new schedule is then transferred to the peripheral devices for implementation and the new schedule is transferred to the optical matrix core scheduler. Implementation of the new schedule as the schedule on the peripheral devices and the optical matrix core scheduler is then performed.

Abstract: Embodiments of the present invention provide a system and method for providing non-blocking routing of optical data through an optical switch fabric. The optical switch fabric can include an optical switching matrix with a plurality of inputs intersecting with a plurality of outputs. A path switch can be located at each intersection that is operable to switch data arriving on an input to a particular output. The path switches can be configurable to create a plurality of unique paths through the optical switching matrix to allow routing in a non-blocking manner. Another aspect of the present invention can provide a system and method for providing non-blocking routing through an optical cross-bar switch. The optical cross-bar switch includes a plurality of input links, a plurality of output links and a plurality of switching elements.

Abstract: An optical Internet Protocol switching method and system are disclosed for switching data packets entirely in the optical domain. The method of the present invention includes the steps of receiving a plurality of data packets at an optical switch and extracting the header information for each data packet for processing at a control unit. The data packets can be demultiplexed at a plurality of demultiplexers and each data packet assigned a different internal wavelength. The method of the present invention can route each data packet to one or more delay buffers based on a current output status. The data packets can be combined into a single output from the delay buffers and broadcast to a plurality of outputs channels. The method of the present invention can select one or more data packets to output through at least one of the output channels and can convert the assigned internal wavelengths of the selected data packets to their original wavelength using a tunable wavelength converter.

Abstract: A switch architecture and method for OCDMA packet switching within a network are disclosed.

Abstract: A scalable Wave Division Multiplexing (WDM) optical IP switching method and system are disclosed for switching data packet payloads in the optical domain.