Abstract: An output of a vision-language model (VLM) can be steered by receiving an input that includes both an image and text, and compositing the image with a universal image configured to elicit specific model activations associated with a targeted behavioral change. The modified image, together with the text, is then supplied to the VLM, resulting in an output that reflects the desired behavioral change. This technique enables steering of the VLM's output without modifying its internal states. The generated output is then provided to an application, process, or system for further utilization. Related apparatus, systems, techniques and articles are also described.

Abstract: The apparatus of an edge computing node, a system, a method and a machine-readable medium. The apparatus includes a processor to perform rounds of federated machine learning training including: processing client reports from a plurality of clients of the edge computing network; selecting a candidate set of clients from the plurality of clients for an epoch of the federated machine learning training; causing a global model to be sent to the candidate set of clients; and performing the federated machine learning training on the candidate set of clients. The processor may perform rounds of federated machine learning training including: obtaining coded training data from each of the selected clients; and performing machine learning training on the coded training data.

Abstract: Techniques for assessing multi-modal inputs to a machine learning model involve receiving a multimodal input containing an image, producing several transformed versions of that image, and generating embeddings for both the original and transformed images. A pairwise similarity analysis among all embeddings is conducted to determine distance values. Two dissimilarity metrics can then be calculated: one reflecting the differences among the transformed images, and another comparing the original image to its transformed versions. If the dissimilarity among the transformed images is greater than that between the original and transformed images plus a threshold, the system triggers a remediation action. This action either blocks the input from being processed by the machine learning model or prevents the model's output from being returned to the requester, thereby enhancing the reliability and security of the model.

Abstract: An apparatus of a transmitter computing node n (TX node n) of a wireless network, one or more computer readable media, a system, and a method.

Abstract: An apparatus of a transmitter computing node n (TX node n) of a wireless network, one or more computer readable media, a system, and a method.

Abstract: Systems and techniques for cross-layer automated fault tracking and anomaly detection are described herein. Anomaly data may be obtained from a plurality of layers of a network. Elements of the anomaly data may be identified that correspond to a data flow of an application executing on the network. An artificial intelligence model may be trained using the elements of the anomaly data to generate an impact score for the application. The impact score may be generated for the application by evaluating current network metrics using the artificial intelligence model. An operational component of the network may be modified based on the impact score.

Abstract: An apparatus may include a trusted execution environment and a processor configured to execute a machine learning (ML)-based application within the trusted execution environment, the ML-based application is configured to provide an output based on input data comprising telemetry data and decrypt encrypted data received by the trusted execution environment to obtain the telemetry data of the network.

Abstract: A plurality of queries are input into an artificial intelligence (AI) model. The AI model is made up of a plurality of layers including an input layer, an output layer, and at least one intermediate layer between the input layer and the output layer. Each intermediate layer, during inference, can output a plurality of activations. Thereafter, for each query, activations are intercepted from at least one of the intermediate layers. It is then determined whether a distribution of the intercepted activations across the queries indicates that the queries seek to cause the AI model to behave in an undesired manner by conducting a distance-based similarity analysis between the intercepted activations and reference activations. Data characterizing such determination is then provided to a consuming application or process. Related apparatus, systems, techniques and articles are also described.

Abstract: In one embodiment, an apparatus of an access point (AP) node of a network includes an interconnect interface to connect the apparatus to one or more components of the AP node and a processor to: access scheduling requests from a plurality of devices, select a subset of the devices for scheduling of resource blocks in a time slot, and schedule wireless resource blocks in the time slot for the subset of devices using a neural network (NN) trained via deep reinforcement learning (DRL).

Abstract: An apparatus of an edge computing node, a method, and a machine-readable storage medium. The apparatus is to decode messages from a plurality of clients within the edge computing network, the messages including respective coded data for respective ones of the plurality of clients; computing estimates of metrics related to a global model for federated learning using the coded data, the metrics including a gradient on the coded data; use the metrics to update the global model to generate an updated global model, wherein the edge computing node is to update the global model by calculating the gradient on the coded data based on a linear fit of the global model to estimated labels from the federated learning; and send a message including the updated global model for transmission to at least some of the clients.

Abstract: A method of implementing distributed AI or ML learning for autonomous vehicles is disclosed. An AI or ML model specific to a location or a type of the location is generated at a vehicle. In response to a detection that the vehicle is within a proximity to a road side unit (RSU) associated with the location or the type of the location or the vehicle is within a proximity to an additional vehicle that is present or anticipated to be present at the location or the type of the location, causing an AI or ML model transmission to the additional vehicle or the RSU. Based on the causing the AI or ML model reception, causing deployment of an additional AI or ML model in the vehicle to optimize the vehicle for the location or the type of the location.

Abstract: Methods, systems, and computer programs are presented for implementing Personalized Mobility as a Service (PMaaS) to improve transportation services delivery. One storage medium includes instructions for detecting, by a mobility as a service (MaaS) system, a request for a trip from a user device of a user. The storage medium further includes instructions for mapping, using a model executing on the machine, the user to a persona from a plurality of persona models. Each persona model has one or more characteristics associated with users of the MaaS system. Further yet, the storage medium includes instructions for determining trip parameters for the trip based on the persona mapped to the user, the trip parameters defining one or more trip segments for the trip, and instructions for providing trip parameters to the user device.

Abstract: Methods, apparatus, systems, and articles of manufacture are disclosed that coordinate network traffic between a wireless network device and a computing platform. An example apparatus includes a wake-up selector to generate a target wait time parameter based on a workload type of a number of packets obtained from a network device and a user preference, the target wait time parameter indicative of a time interval that, when met, causes a modem to retrieve the number of packets, a data frame generator to generate a data frame that causes the network device to buffer the number of packets for the time interval, and a network packet controller to negotiate, using the data frame, the target wait time parameter with a network device.

Abstract: A method can be performed by a first node for determining a parameter of physical (PHY) layer circuitry of a second node. The method can include implementing a cascaded hierarchy of techniques to determine, based on an electrical signal from a second node, a parameter of the PHY layer circuitry of the second node, and causing an antenna of the first node to transmit an electromagnetic wave consistent with the determined parameter.

Abstract: Methods, apparatus, systems, and articles of manufacture are disclosed that generate dynamic latency values. An example apparatus includes an active status controller to determine that a modem is active based on a number of packets obtained from a network, a prediction controller to predict that the number of packets are indicative of a workload type based on a trained model, and a latency value generator to generate a latency value based on the workload type of the number of packets, the latency value to cause a processor processing the number of packets to enter a power saving state or a power executing state.

Abstract: System and techniques for information centric network (ICN) protocol for federated learning are described herein. An interest packet may be received on a first interface to start a federated learning round. Here, the interest packet includes a participant criterion and a federated learning round expiration. An entry, that includes the federated learning round expiration, is created in a pending interest table (PIT) for the interest packet. The interest packet is forwarded, in accordance with a forwarding information base (FIB), to a set of interfaces before the federated learning round expiration. When a data packet from a node, that meeting the participant criterion, is received in response to the interest packet, the data packet is forwarded on the first interface in accordance with the PIT entry.

Abstract: An apparatus of an edge computing node, a method, and a machine-readable storage medium. The apparatus is to decode messages from a plurality of clients within the edge computing network, the messages including respective coded data for respective ones of the plurality of clients; computing estimates of metrics related to a global model for federated learning using the coded data, the metrics including a gradient on the coded data; use the metrics to update the global model to generate an updated global model, wherein the edge computing node is to update the global model by calculating the gradient on the coded data based on a linear fit of the global model to estimated labels from the federated learning; and send a message including the updated global model for transmission to at least some of the clients.

Abstract: Systems and techniques for information centric network (ICN) interworking are described herein. For example, a request may be received at a convergence layer of a node. Here, the request originates from an application on the node. A network protocol, from several available to the node, may be determined to transmit the request. The node then transmits the request via the selected network protocol.

Abstract: Techniques for non-linear distributed multitask support vector machines are disclosed. In the illustrative embodiment, a coordinator node sends initial parameters (or a random number generator along with model choice) for a global model to participant nodes. Each participant node performs a round of training based on the common global model parameters, the model models, and local data. Each participant node determines updated parameters for the global model and updated parameters for a local model. Each participant node sends an update of the parameters to the global model to the coordinator node, while keeping the parameters of the local model private. The coordinator node aggregates the updates from the participant nodes, updates the global model parameters, and sends them back to the participant nodes. The process can repeat until a desired error level is reached.

Abstract: A task modeling system, including a plurality of processing clients having a plurality of processing cores; a task modeler, including a memory storing an artificial neural network; and a processor, configured to receive input data representing a plurality of processing tasks to be completed by the processing client within a predefined time duration; and implement its artificial neural network to determine from the input data an assignment of the processing tasks among the processing cores for completion of the processing tasks within the predefined time duration, and determine a power management factor for each of the plurality of processing cores for power management during the predefined time duration; wherein the artificial neural network is configured to select the power management factor for each of the plurality of processing cores to achieve a power usage within a predefined threshold for the plurality of processing cores during the predefined time duration.