Abstract: A computer-implemented method includes receiving an anomaly clustering request that requests data processing hardware to assign each image of a plurality of images into one of a plurality of groups. The method also includes obtaining a plurality of images. For each respective image, the method includes extracting a respective set of patch embeddings from the respective image, determining a distance between the respective set of patch embeddings and each other set of patch embeddings, and assigning the respective image into one of the plurality of groups using the distances between the respective set of patch embeddings and each other set of patch embeddings.

Abstract: A method for sharing assistant profiles includes receiving, at a profile service, from an assistant service interacting with a user device of a user, a request requesting the profile service to release personal information associated with the user to the assistant service. The operations also include performing, through the assistant service, a verification process to verify that the user consents to releasing the requested personal information by: instructing the assistant service to prompt the user to recite a unique token prescribed to the user; receiving audio data characterizing a spoken utterance captured by the user device of the user; processing the audio data to determine whether a transcription of the spoken utterance recites the unique token; and when the transcription of the spoken utterance recites the unique token, releasing, to the assistant service, the requested personal information stored on a centralized data store managed by the profile service.

Abstract: A method for anomaly detection includes receiving an anomaly detection query from a user. The anomaly detection query requests data processing hardware determine one or more anomalies in a dataset including a plurality of examples. Each example in the plurality of examples is associated with one or more features. The method includes training a model using the dataset. The trained model is configured to use a local outlier factor (LOF) algorithm. For each respective example of the plurality of examples in the dataset, the method includes determining, using the trained model, a respective local deviation score based on the one or more features. The method includes determining that the respective local deviation score satisfies a deviation score threshold and, based on the location deviation score satisfying the threshold, determining that the respective example is anomalous. The method includes reporting the respective anomalous example to the user.

Abstract: Secure methods, systems, and media for generating and verifying user credentials are provided. In some embodiments, the method comprises: receiving, from a user device, a request for access to a service that requires valid user credentials; determining an aspect of the user credentials that is to be satisfied to grant access to the requested service; transmitting, to the user device, a request for information related to the aspect of the user credential; receiving, from the user device, information related to the aspect of the user credential, wherein the information has been signed using a key associated with the user device; verifying the key used to sign the information by the user device; in response to verifying the key used to sign the information, determining whether the aspect of the user credential has been satisfied based on the received information; and, in response to determining that the aspect of the user credential has been satisfied, granting access to the service.

Abstract: A method for communicating location information to a device includes receiving, at a computer system that implements a social networking service, location information that represents a geographic location of a device associated with a first user; associating, by the computer system, the received location information with a profile associated with the first user; and sending, from the computer system to a device associated with a second user, a message that is generated based at least in part on the location information.

Abstract: An automated rack imaging system is provided, including an automated guided vehicle having a housing and a propulsion system configured to move the housing. The automated rack imaging system may include an imaging system coupled to the housing. The imaging system may include a plurality of cameras. The cameras each may be configured to have a respective field of view. The fields of view may be at least partially non-overlapping with one another. The automated rack imaging system may also include an image processor configured to combine a plurality of images taken by the cameras into a single mosaic image. A method of imaging a datacenter rack with the automated guided vehicle is also provided. The method may include moving the automated guided vehicle to a first target location aligned with the datacenter rack, taking the plurality of images, and combining the plurality of images into the single mosaic image.

Abstract: A method for storing serialized structured data generically in a standardized serialized data structure. The method includes obtaining structured data that includes one or more field pairs and transcoding the structured data into serialized self-describing data. Each field pair includes a corresponding field identifier and a field value associated with the corresponding field identifier. The serialized self-describing data includes one or more self-describing data portions each representing a corresponding one of the one or more field pairs. Each self-describing portion of the one or more self-describing portions includes a first series of bits representing the corresponding field identifier, and a second series of bits representing the field value associated with the corresponding field identifier. The method also includes transmitting the serialized self-describing data to a remote entity.

Abstract: Methods, systems, and apparatus, including computer-readable media, are described for performing neural network computations using a system configured to implement a neural network on a hardware circuit. The system includes a host that receives a batch of inputs to a neural network layer. Each of the inputs is stored in a memory location identified by an address. The system identifies one or more duplicate addresses in a listing of addresses for one or more inputs. For each duplicate address: the system generates a unique identifier that identifies the duplicate address in the listing of addresses. The system (i) obtains first inputs from memory locations identified by addresses corresponding to the unique identifiers and (ii) generates an output of the layer from the obtained first inputs.

Abstract: Heat dissipation and electric shielding techniques and apparatuses are disclosed to enable the operation of OSFP modules at higher bandwidths. OSFP compatible techniques are discussed including the use of water cooling, addition of heat pipes, use of intercoolers, air-fins and air-foils, optimization of cooling fins, use of vapor chambers are discussed.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for obtaining i) historical location data specifying geographic locations of the user over a period of time that is included in a user profile of the user and ii) predefined geographic location data specifying a predefined geographic location of the user that is included in the user profile of the user; identifying a pair of cross-contaminated geographic regions that both include geographic locations specified by the historical location data of the user profiles for each of a threshold number of users of the plurality of users, and in response, merging the pair of cross-contaminated geographic regions to define a merged geographic region; and merging the merged geographic region with additional geographic regions until a cross-contamination between a resulting merged geographic region created by the merging and other geographic regions is reduced to a specified level of cross-contamination.

Abstract: Methods, systems, and apparatus for nonlinear calibration of quantum computing apparatus. In one aspect, elements in a set of experimental data correspond to a respective configuration of control biases for the quantum computing apparatus. An initial physical model comprising one or more model parameters of the quantum computing apparatus is defined. The model is iteratively adjusted to determine a revised physical model, where at each iteration: a set of predictive data corresponding to the set of experimental data is generated, and elements in the predictive data represent a difference between the two smallest eigenvalues of a Hamiltonian characterizing the system qubits for the previous iteration, and are dependent on at least one model parameter of the physical model for the previous iteration; and the model for the previous iteration is adjusted using the obtained experimental data and the generated set of predictive data for the iteration.

Abstract: Methods, systems, and apparatus for parallel optimization of continuously running quantum error correction by closed-loop feedback. In one aspect, a method includes continuously and effectively optimizing qubit performance in-situ whilst an error correction operation on the quantum system is running. The method directly monitors the output from error detection and provides this information as feedback to calibrate the quantum gates associated with the quantum system. In some implementations, the physical qubits are spatially partitioned into one or more independent hardware patterns, where the errors attributable to each hardware pattern are non-overlapping. The one or more different sets of hardware patterns are then temporarily interleaved such that all physical qubits and operations are optimized. The method allows for the optimization of each section of a hardware pattern to be performed individually and in parallel, and can result is O(1) scaling.

Abstract: An application specific integrated circuit (ASIC) chip includes: a systolic array of cells; and multiple controllable bus lines configured to convey data among the systolic array of cells, in which the systolic array of cells is arranged in multiple tiles, each tile of the multiple tiles including 1) a corresponding sub array of cells of the systolic array of cells, 2) a corresponding subset of controllable bus lines of the multiple controllable bus lines, and 3) memory coupled to the subarray of cells.

Abstract: Systems and methods of error handling in a network interface card (NIC) are provided. For a data packet destined for a local virtual machine (VM), if the NIC cannot determine a valid translation memory address for a virtual memory address in a buffer descriptor from a receive queue of the VM, the NIC can retrieve a backup buffer descriptor from a hypervisor queue, and store the packet in a host memory location indicated by an address in the backup buffer descriptor. For a transmission request from a local VM, if the NIC cannot determine a valid translated address for a virtual memory address in the packet descriptor from a transmit queue of the VM, the NIC can send a message to a hypervisor backup queue, and generate and transmit a data packet based on data in a memory page reallocated by the hypervisor.

Abstract: A computer-implemented method executed by one or more processors, the method includes monitoring execution of program code executed by a first processor component; and monitoring execution of program code executed by a second processor component. A computing system stores data identifying hardware events in a memory buffer. The stored events occur across processor units that include at least the first and second processor components. The hardware events each include an event time stamp and metadata characterizing the event. The system generates a data structure identifying the hardware events. The data structure arranges the events in a time ordered sequence and associates events with at least the first or second processor components. The system stores the data structure in a memory bank of a host device and uses the data structure to analyze performance of the program code executed by the first or second processor components.

Abstract: A smart-home device may include an energy-storage element that stores energy harvested from an environmental system; a power wire connector and a return wire connector; and switching elements configured to operate in a first state where the switching elements create a connection between the power and the return; and a second state where the switching elements interrupt the connection between the power and return. The smart-home device may also include a circuit that controls the switching elements, where the circuit is configured to detect a zero-crossing of a current received through the power wire connector; wait for a first time interval after the zero-crossing is detected; after an expiration of the first time interval, enable active power stealing for a second time interval; and after an expiration of the second time interval, disable active power stealing.

Abstract: Biquad stage systems and methods include receiving at biquad sections a signal sample, generating, by each biquad section, a pair of output values based on the signal sample, including a first value based on fixed-point processing path and a second value emulating a floating-point processing path, and accumulating the pair of output values from each of the plurality of biquad sections to generate an output signal. The biquad stage receives an N-bit input signal, which is processed by a biquad section. Delay elements delay the signal sample before input to other biquad sections. The delayed signal sample is input to the first processing path and the second processing path of a corresponding biquad stage. By performing the processing based on two paths, a more accurate result can be found when using a reduced word length in the multiply operations resulting in a lowering of the power consumption.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for adversarial training of a neural network. One of the methods includes obtaining a plurality of training inputs; and training the neural network on each of the training inputs, comprising, for each of the training inputs: processing the training input using the neural network to determine a neural network output for the training input; applying a perturbation to the training input to generate an adversarial perturbation of the training input; processing the adversarial perturbation of the training input using the neural network to determine a neural network output for the adversarial perturbation; and adjusting the current values of the parameters of the neural network by performing an iteration of a neural network training procedure to optimize an adversarial objective function.

Abstract: A method performs large-scale data processing in a distributed and parallel processing environment. The method defines application-independent map and reduce operations, each invoking one or more library functions that automatically handle data partitioning, parallelization of computations, and fault tolerance. A user specifies a map operation, which calls one or more of the application-independent map operators to perform data read and write operations. A user also specifies a reduce operation, which calls one or more of the application-independent reduce operators to perform data read and write operations. The method executes application-independent map worker processes. Each map worker process executes the user-specified map operation to read designated portions of input files and store intermediate data values in intermediate data structures. The method also executes application-independent reduce worker processes.

Abstract: A method of processing messages executes at a computing device having one or more processors and memory. The memory stores one or more programs configured for execution by the one or more processors. A first message for a user is analyzed for a structured content element. When found, content from a data source distinct from the message is obtained by executing an associated action. A first message display state is formed for the message comprising a message notification and the content. A messaging application user interface, comprising an electronic message list with a plurality of objects, is updated. Responsive to selection of a first object of the plurality of objects, the first object representing the first message, the user is enabled to toggle the first message display state between first and second display states, the second display state differing from the first by providing direct access to the first message.