Abstract: Transmission Control Protocol (TCP) states of an active device are replicated at a backup device configured to track connections in a software defined network (SDN). The backup device receives, from the active device, a TCP packet with a TCP flag including one or more of SYN, SYN-ACK, ACK, FIN, FIN-ACK, ACK, or RESET. When the TCP packet has a SYN flag, the backup device adds a connection record to a connection table. Otherwise, the backup device derives an updated connection state for the connection record using a TCP state machine. The connection state is updated as future packets are received by the backup device.

Abstract: Data flows in a virtualized computing environment are efficiently updated by a hardware-based networking device configured to disaggregate processing of data packets of the data flows from hosts of the virtualized computing environment. A connection table is accessed that defines connection flows for data packets having a source from an endpoint in a virtual network of the virtualized computing environment or a destination to the endpoint in the virtual network of the virtualized computing environment. The hardware-based networking device re-simulates full packet processing paths for each of the flows in the connection table and updates the flows in the connection table to ensure that the flows in the connection table implement policies of the virtualized computing environment that were updated after corresponding flows in the connection table were added to the connection table.

Abstract: As opposed to a rigid approach, implementations disclosed herein utilize a flexible approach in automatically determining an action set to utilize in attempting performance of a task that is requested by natural language input of a user. The approach is flexible at least in that embedding technique(s) and/or action model(s), that are utilized in generating action set(s) from which the action set to utilize is determined, are at least selectively varied. Put another way, implementations leverage a framework via which different embedding technique(s) and/or different action model(s) can at least selectively be utilized in generating different candidate action sets for given NL input of a user. Further, one of those action sets can be selected for actual use in attempting real-world performance of a given task reflected by the given NL input. The selection can be based on a suitability metric for the selected action set and/or other considerations.

Abstract: This disclosure relates to improved systems, methods, and techniques for constructing and employing neural network architectures to solve computer vision and other problems. The neural network architectures can have two or three layers with all nodes in the first layer connected to all nodes in the second layer. The nodes in the second layer can be connected to each other. The weights or values of the various connections between these nodes in the first two layers can also be updated between the processing of inputs to the neural network architectures. These neural network architectures do not require extensive training and can learn continuously. Other embodiments are disclosed herein as well.

Abstract: This disclosure relates to improved systems, methods, and techniques for constructing and employing neural network architectures to solve computer vision and other problems. The neural network architectures can have two or three layers with all nodes in the first layer connected to all nodes in the second layer. The nodes in the second layer can be connected to each other. The weights or values of the various connections between these nodes in the first two layers can also be updated between the processing of inputs to the neural network architectures. These neural network architectures do not require extensive training and can learn continuously. Other embodiments are disclosed herein as well.

Abstract: In implementations of three-dimensional vector-based brushes, a computing device implements a brush system to receive input data describing a first stamp and a second stamp of consecutive stamps of a vector-based digital brush. The first stamp is segmented into a first convex geometry and a second convex geometry, and the second stamp is segmented into a third convex geometry and a fourth convex geometry. The brush system computes a first convex hull of the first convex geometry and the third convex geometry and computes a second convex hull of the second convex geometry and the fourth convex geometry. An order for passing the first convex hull and the second convex hull to a planarizer is determined and the brush system generates a vector-based stroke of digital paint for display in a user interface based on the order.

Abstract: Computer-implemented methods for rendering a volumetric dataset and a surface embedded in the volumetric dataset are described. One method includes performing a volume rendering process to generate a volume rendering of the volume. Based on the volume rendering process, depths of respective locations in the volume rendering are determined, and the depths stored in association with the respective locations. A surface rendering process generates a surface rendering of a surface using the depths and respective locations. The volume rendering and the surface rendering are combined into a combined rendering of the volume and surface. In another method, depths of the surface at respective locations with respect to the volume are determined and the depths are stored in association with the locations. A volume rendering process includes a shading process which uses the depths of the surface and the respective locations.

Abstract: Stacked transistor structures having a conductive interconnect between source/drain regions of upper and lower transistors. In some embodiments, the interconnect is provided, at least in part, by highly doped epitaxial material deposited in the upper transistor’s source/drain region. In such cases, the epitaxial material seeds off of an exposed portion of semiconductor material of or adjacent to the upper transistor’s channel region and extends downward into a recess that exposes the lower transistor’s source/drain contact structure. The epitaxial source/drain material directly contacts the lower transistor’s source/drain contact structure, to provide the interconnect. In other embodiments, the epitaxial material still seeds off the exposed semiconductor material of or proximate to the channel region and extends downward into the recess, but need not contact the lower contact structure.

Abstract: Techniques are described herein for translating source code using sparse-self attention. In various implementations, a source code snippet in a first programming language may be processed to obtain graph(s) representing snippet tokens, and relationships therebetween. Based on the graph(s), a subset of snippet token pairs may be identified from a superset of all possible token pairs in the source code snippet. Each token pair of the subset may include snippet tokens that are represented by nodes connected by one or more edges of the one or more graphs. A self-attention network of a translation machine learning model may be adapted to sparsely attend across the identified subset of token pairs. The source code snippet may then be processed based on the adapted translation machine learning model to generate a translation of the source code snippet in the second programming language.

Abstract: The present technology can move operating system folders into a sync folder of a cross platform content management system, and redirect the operating system to look for the OS folders in the sync folder. The present technology also provides an invariant checker to make sure that another application has not moved the OS folders after they have been placed in the sync folder, and provides solutions when the OS folders are moved out of the sync folder of the content management system. Additionally, when OS folders for multiple client devices are in the sync folder on the content management system, the present technology can provide a mechanism to make the content items in an OS folder on a first client device also sync into an OS folder on second client device.

Abstract: A system and method for automatic optimization of user engagement on a digital platform is provided. The method comprises generating, a set of tokens for a gamified challenge on the digital platform. The method thereafter leads to associating token(s) from the set of tokens with a set of levels of the gamified challenge. Further the method encompasses providing to user(s) via the digital platform, action item(s) at least to enable the user(s) to retrieve token(s) from the set of tokens and to permit the user(s) to transition to a next level from a current level. Further the method comprises determining for each level, a probability of retrieving the token(s) by the user(s). The method thereafter encompasses automatically controlling, a number of users permitted to transition to the next level from the current level based on the determined probability, to automatically optimize the user engagement on the digital platform.

Abstract: A client of a distributed database service may conveniently enable replication of data from a source table to a destination (e.g., a different service/database table) by submitting a request enable replication for the source table. In response to the request to enable replication, the database service enables replication for the source table. To replicate a change, the database service may modify a format of the change based on a data storage format of the destination that is different than a data storage format of the database service. The database service monitors the performance of the replication and based on the performance, the database service may automatically modify the capacity to perform the replication of changes to the destination (e.g., by assigning additional resources to replication).

Abstract: A serverless computing system is configured to provide access to a machine learning model by at least associating an endpoint, comprising code that accesses the machine learning model, with an extension that interfaces between a serverless compute architecture and the endpoint. A request to perform an inference is received by the system and processed by using the serverless compute architecture to execute a compute function. The compute function cases the extension to interface with the endpoint to cause the machine learning model to perform the inference.

Abstract: Embodiments described herein provide a soft prompt tuning technique referred to as the Vector quantized Input-contextualized Prompt (VIP). The VIP techniques has two integral properties i) instead of learning a fixed set of prompt tokens irrespective of the input, it generates a contextualized version of the soft prompts, conditional on the input text ii) it further passes the input-contextualized prompt tokens through a quantization network, inspired by Vector Quantized Transformers. The quantization network uses nearest neighbor search over a learnable codebook to train a discrete latent variable model over the prompt-space, thus generating quantized version of contextual prompt tokens. These quantized contextual prompt tokens are finally fed into the frozen language model along with the original input text.

Abstract: Techniques for processing web probes for monitoring user experience including use of caching to prevent a surge of web probes on destination servers and for detecting web probe traffic. A method implemented by a connector includes intercepting a Hypertext Transfer Protocol Secure (HTTPS) web probe request to a server, identifying a cache hit associated with the request in a cache, generating a synthetic Hypertext Transfer Protocol (HTTP) response based on information from the identified cache hit, wherein the generated synthetic HTTP response includes an extension header containing collected statistics, and sending the synthetic HTTP response. The method can further include simulating a Secure Socket Layer (SSL) handshake to estimate SSL cost.

Abstract: Embodiments described herein address these and other issues by providing radio frequency (RF) sensing to determine the status of a driver or other occupant of the vehicle. RF sensing may be provided by existing radios of a vehicle, such a Wi-Fi transceiver, and may therefore provide RF sensing functionality to a vehicle with little added cost. RF sensing can be leveraged to implement safety features such as detecting an unattended child or pet in a vehicle, detecting driver alertness, and the like.

Abstract: Described is a method of frame-wise encoding metadata for an input signal, the metadata comprising a plurality of at least partially interrelated parameters calculable from the input signal. The method comprises, for each frame: iteratively performing, by using a looping process, steps of: determining a processing strategy from a plurality of processing strategies for calculating and quantizing the parameters; calculating and quantizing the parameters based on the determined processing strategy to obtain quantized parameters; and encoding the quantized parameters. In particular, each of the plurality of processing strategies comprises a respective first indication indicative of an ordering related to the calculation and quantization of individual parameters; and the processing strategy is determined based on at least one bitrate threshold.

Abstract: Embodiments described herein provide a soft prompt tuning technique referred to as the Vector quantized Input-contextualized Prompt (VIP). The VIP techniques has two integral properties i) instead of learning a fixed set of prompt tokens irrespective of the input, it generates a contextualized version of the soft prompts, conditional on the input text ii) it further passes the input-contextualized prompt tokens through a quantization network, inspired by Vector Quantized Transformers. The quantization network uses nearest neighbor search over a learnable codebook to train a discrete latent variable model over the prompt-space, thus generating quantized version of contextual prompt tokens. These quantized contextual prompt tokens are finally fed into the frozen language model along with the original input text.

Abstract: Techniques are disclosed for processing data packets and implementing policies in a software defined network (SDN) of a virtual computing environment. At least two SDN appliances are configured to disaggregate enforcement of policies of the SDN from hosts of the virtual computing environment. The hosts are implemented on servers communicatively coupled to network interfaces of the SDN appliance. The servers host a plurality of virtual machines. The servers are communicatively coupled to network interfaces of at least two top-of-rack switches (ToRs). The SDN appliance comprises a plurality of smart network interface cards (sNICs) configured to implement functionality of the SDN appliance. The sNICs have a floating network interface configured to provide a virtual port connection to an endpoint within a virtual network of the virtual computing environment.

Abstract: Embodiments herein describe techniques for a semiconductor device including a carrier wafer, and an integrated circuit (IC) formed on a device wafer bonded to the carrier wafer. The IC includes a front end layer having one or more transistors at front end of the device wafer, and a back end layer having a metal interconnect coupled to the one or more transistors. One or more gaps may be formed by removing components of the one or more transistors. Furthermore, the IC includes a capping layer at backside of the device wafer next to the front end layer of the device wafer, filling at least partially the one or more gaps of the front end layer. Moreover, the IC includes one or more air gaps formed within the one or more gaps, and between the capping layer and the back end layer. Other embodiments may be described and/or claimed.