Dell Products Inc.

Emc IP Holding Company LLC

Nadav Azaria

Ofir Ezrielev

Avitan Gefen

Jehuda Shemer

Saar Cohen

Amihai Savir

Yonit Weiss

Methods and systems for anomaly detection in a distributed environment are disclosed. To manage anomaly detection, a system may include an anomaly detector and one or more data collectors. The anomaly detector may detect anomalies in data obtained from one or more of the data collectors using an inference model and an anomaly threshold. The anomaly threshold may be determined by fitting a normal distribution to output of the inference model when the inference model is exercised across an input range of the inference model. The anomaly threshold may correspond to a portion of the normal distribution. To perform anomaly detection, the inference model may require periodic re-training. Data collected from the one or more data collectors may be used to re-train the inference model as needed. Following anomaly detection and/or inference model re-training, the data may be discarded to remove the data from the anomaly detector.

System and method for memory-less anomaly detection using anomaly thresholds based on probabilities

Methods and systems for anomaly detection in a distributed system are disclosed. To manage anomaly detection, a system may include an anomaly detector and one or more data collectors. The anomaly detector may detect anomalies in data obtained from one or more of the data collectors using an inference model. The inference model may be an autoencoder trained to reconstruct data that is intended to match input data to an extent considered acceptable by the system. To accurately perform anomaly detection, the inference model may require re-training. Data collected from the one or more data collectors may be used to re-train the inference model as needed. Following anomaly detection and/or inference model re-training, the data may be discarded to remove the data from the anomaly detector.

System and method for memory-less anomaly detection using an autoencoder

In one or more embodiments, one or more systems, one or more methods, and/or one or more processes may execute a process; provide input data to the process as the process executes; receive output data from the process as the process executes; after executing the process, determine a neural network based at least on the input data to the process and the output data from the process; determine multiple binary neural networks from the neural network; determine a network of multiple logic gates based at least on the multiple binary neural networks of the neural network; and configure an integrated circuit based at least on the network of the multiple logic gates. For example, the integrated circuit may include at least one of a field programmable gate array, an application specific integrated circuit, a programmable array logic, and a complex logic device.

System and method of configuring integrated circuits

One example method includes receiving, at an IO journal, a new entry that identifies a respective disk location L, and data X written at that disk location L, and determining whether a location specified in an oldest entry of the IO journal is specified in any other entries in the IO journal. When the location specified in the oldest entry is not specified in any other entries in the IO journal, adding the new entry to the IO journal, and augmenting the new entry with undo data. Or, when the location specified in the oldest entry is specified in at least one other entry in the journal, setting data specified in the oldest entry as undo data for the next entry that identifies that location, and adding the new entry to the IO journal, and deleting the oldest entry from the IO journal.

System and a method for generating live VM images with an extended recovery range

Methods and systems for inference generation are disclosed. To manage inference generation, a system may include an inference model manager and any number of data processing systems. The inference model manager may partition an inference model into portions. Portions of the inference model may be distributed to data processing systems in accordance with an execution plan. The execution plan may include instructions for timely execution of the inference model with respect to the needs of a downstream consumer. The inference model manager may manage execution of the inference model by monitoring the functionality of the data processing systems and dynamically re-assigning and/or re-locating data processing systems in the event that one or more data processing systems becomes unable to execute a portion of the inference model.

System and method for inference generation through dynamic reassignment of inference model portions

Methods and systems for inference generation are disclosed. To manage inference generation, a system may include an inference model manager and any number of data processing systems. The inference model manager may represent an inference model as a bipartite graph. To obtain portions of the inference model, the bipartite graph may be partitioned into portions and the portions may undergo an optimization process. The optimization process may include adding and/or competing for nodes of the bipartite graph in order to increase the stability of the portion with respect to the available computing resources of a corresponding data processing system. The optimization process may continue until all portions achieve stability in a way that reduces necessary communications between the data processing systems. Each portion of the inference model may be distributed to one data processing system so that the data processing systems may collectively generate inferences usable by a downstream consumer.

System and method for inference generation via optimization of inference model portions

Methods and systems for inference generation are disclosed. To manage inference generation, a system may include an inference model manager and any number of data processing systems. The inference model manager may represent an inference model as a bipartite graph in order to obtain portions of the inference model. Each portion of the inference model may be distributed to one data processing system so that the data processing systems may collectively generate inferences usable by a downstream consumer. Portions of the inference model may be obtained so that each portion matches the available computing resources of a data processing system throughout the distributed environment. In addition, the portions may be obtained in order to reduce inter-data processing system communications during execution of the inference model.

System and method for execution of inference models across multiple data processing systems

Methods and systems for anomaly detection in a distributed environment are disclosed. To manage anomaly detection, a system may include an anomaly detector and one or more data collectors. The anomaly detector may detect anomalies in data and classify the anomalies based on magnitudes of anomalies using an inference model. Different magnitudes of anomalies may be keyed to different action sets in response to the presence of anomalies in data. To perform anomaly detection, the inference model may require re-training. Data collected from the one or more data collectors may be used to re-train the inference model as needed. Following anomaly detection and/or inference model re-training, the data may be discarded to remove the data from the anomaly detector.

System and method for memory-less anomaly detection using anomaly levels

Techniques are provided for adaptive sensor position determination for multiple mobile sensors. One method comprises obtaining a spatio-temporal representation of sensor measurements, from multiple mobile sensors, wherein the spatio-temporal representation comprises multiple layers each corresponding to a different point in time, wherein a given layer comprises multiple positions, and wherein each position in the given layer corresponds to a possible location for at least one of the multiple mobile sensors in an environment; applying the spatio-temporal representation to an environment state prediction model that generates a prediction of at least one future sensor measurement value for multiple positions in the spatio-temporal representation; applying the predictions of the at least one future sensor measurement value to a sensor position determination model that determines a new position for each of one or more of the multiple mobile sensors; and initiating a movement of the one or more of the multiple mobile sensors to the new position.

Adaptive sensor position determination for multiple mobile sensors

A system can divide a database into a group of shards that are distributed among a group of data centers. The system can train a machine learning model on a group of labeled input data, wherein the group of labeled input data comprises respective requests to operate on the database, and wherein the respective requests are labeled with respective shards of the group of shards used to process the respective requests, and to produce a trained machine learning model. The system can, after training the machine learning model, receive a request. The system can process the request with the trained machine learning model to predict that a data center of the group of data centers will have a largest number of leader shards of the group of shards to process the request. The system can send the request to the first data center to be processed.

Sharded database load distributor

A method for alleviating data poisoning in an edge computing resource includes receiving a numeric value from an Internet of Things (IoT) unit and associating the numeric value with a cluster selected from a plurality of clusters in accordance with a suitable clustering algorithm such as a k-means clustering algorithm. In at least some embodiments, the numeric value comprises a poisoned numeric value including an adversarial component injected by an adversary to negatively impact a trained model of a cloud-based artificial intelligence engine. Rather than permitting the injected adversarial component to corrupt the AI engine, a cluster with which the numeric value is associated is sampled in accordance with a probability distribution of the cluster to obtain a surrogate for the poisoned numeric value. The surrogate may then be provided as an input to an inference module of the AI engine to generate a prediction.

Cluster resampling for alleviation of data poisoning on the edge

Methods and systems for seamlessly changing over between inference models are disclosed. The inference models may be distributed across multiple data processing systems. Provide a seamless changeover, updated inference models and original inference models may be managed in accordance with an update framework. The update framework may ensure that the original inference model continues to operate until all of the portions of the updated inference model are in place and ready to operate. During the update process, the update framework may ensure that redundancy goals continue to be met so that failures of some of the data processing systems are not be fatal to continued operation of at least one of the inference models, such as the original or updated inference model.

System and method for managing transition between inference models across multiple data processing systems

One example method includes receiving, at an IO journal, a new entry that identifies a respective disk location L, and data X written at that disk location L, and determining whether a location specified in an oldest entry of the IO journal is specified in any other entries in the IO journal. When the location specified in the oldest entry is not specified in any other entries in the IO journal, adding the new entry to the IO journal, and augmenting the new entry with undo data. Or, when the location specified in the oldest entry is specified in at least one other entry in the IO journal, setting data specified in the oldest entry as undo data for the next entry that identifies that location, and adding the new entry to the IO journal, and deleting the oldest entry from the IO journal.

A system can establish a primary file system on a block array of computer storage, wherein the block array comprises a range of addresses, wherein the primary file system is configured to address the range of addresses. The system can establish a shadowed file system on the block array, wherein the shadowed file system is configured to access portions of the block array that are unused by the primary file system. The system can, in response to receiving a request to write data to the primary file system, and in response to determining that an amount of the block array utilized by the primary file system is full, transfer a first portion of the block array utilized by the shadowed file system to the primary file system.

Growing community of inventors

Meitar, Israel

Nadav Azaria