Future Wei Technologies, Inc.

Huawei Technologies Co., Limited

Jin Yang

Yongxi Tan

Ying Li

Baoling S Sheen

Nandu Gopalakrishnan

Ruilin Liu

Yan Xin

Kai Yang

Yanjia Sun

Qitao Song

Frank J Effenberger

Xiang Liu

Ning Cheng

Yanjie Fu

Iyad Alfalujah

James Mathew

Xufeng Chen

Kamalaharan Dushyanthan

James Matthew

Yunjun Chen

Xin Zhang

Zhangxiang Ye

Lizhou Zhou

Ji Wang

Xie Wang

Jie Ren

Xiujun Shu

Wenwen Zhao

Yanchi Liu

Deti Liu

Xin Li

An apparatus comprises a first mirror; a second mirror; a modulation layer positioned between the first mirror and the second mirror and comprising a plurality of modulation regions; a diffraction layer positioned between the modulation layer and the second mirror, and an input port admitting a light beam into the apparatus. The light beam passes through the diffraction layer and is modulated by the modulation layer to create a first modulated beam before being reflected by the first mirror, the first mirror reflecting the first modulated beam toward the second mirror, the second mirror reflecting the first modulated beam toward the modulation layer to be modulated for at least a second time.

Diffractive deep neural network (D2NN) processing using a single modulation layer

The disclosure relates to technology for configuring parameters in a wireless communications network. Parameter configurations resulting in a change to key quality indicator (KQI) and key performance indicator (KPI) measurements are determined based on collected data samples. The data samples are divided into subsets including a first subset including the data samples associated with the parameter configurations failing to result in the change to the KQI and KPI measurements, and a second subset including the data samples associated with the parameter configurations resulting in the change to the KQI and KPI measurements dependent upon satisfying conditions in the wireless communications network. The subsets of the data samples are then determined for using machine learning to optimize the parameter configurations, and subsets of the data samples are provided as an input to machine learning for the parameter configurations to optimize the wireless communications network.

Automatically optimize parameters via machine learning

A computer-implemented method for predicting received signal strength in a telecommunication network includes receiving, by one or more processors that execute a convolutional neural network, geographic data representing geographic information of a geographic area and antenna and transmit power information of a base station in the geographic area; inputting the geographic data and the antenna and transmit power information into the convolutional neural network; predicting received signal strength using the convolutional neural network that includes a number of convolution layers based on the received geographic data and the antenna and transmit power information, the received signal strength representing signal strength of wireless signals received at different locations in the geographic area; and outputting the predicted received signal strength.

Predicting received signal strength in a telecommunication network using deep neural networks

Techniques are provided for monitoring the performance of a user device in a communication network. The techniques include detecting an anomaly in a performance measurement such as a key quality indicator (KQI) of the user device. The techniques include obtaining historical measurements of the KQI for user devices. The historical measurements are assigned to states to reflect whether the performance is good or bad, or somewhere in between. The states can be defined differently for different hours in the day so that the states represent the relative performance for that time of day. For each user device, a Markov model is provided indicating probabilities of transitions between the states. Additional measurements are obtained of the KQI for a selected user device, and the Markov model of the selected user device is used to detect an anomaly in the additional measurements.

User-level KQI anomaly detection using markov chain model

A processor implemented method of identifying a root cause of degraded network quality in a wireless network. The method includes accessing historical network performance data, the performance data including a time sequenced measure of performance indicators for the network. The method evaluates the historical performance data to determine regularly occurring associations between indicators to define a set of rules characterizing the associations of the wireless network, and stores the set of rules in a data structure. The wireless network is monitored by accessing analysis data reporting time sequenced performance indicator data. Next, anomalies are detected in a performance indicator in the analysis data and matched to at least one rule in the set of rules. The method outputs an indication of a cause of degradation in the wireless network resulting from the anomaly in the performance indicator.

Fingerprinting root cause analysis in cellular systems

Techniques for optimizing configuration parameters in a wireless network may iteratively adjust one or more wireless configuration parameters during a first stage of an adjustment period until either a cost increase during a single iteration exceeds an upper cost threshold or (ii) the number of successive iterations exhibiting a per-iteration cost increase between the upper cost threshold and a lower cost threshold exceeds a threshold number of iterations. This may trigger the AP proceed to a second stage of the adjustment period, where the AP may adjust the same wireless configuration parameter in an opposite direction, maintain a value of the wireless parameter, or adjust a different wireless configuration parameter during the second stage. Progression between successive adjustment periods may be at least partially coordinated between APs in a group of APs.

Self-learning, adaptive approach for intelligent analytics-assisted self-organizing-networks (SONs)

A neural network is trained using deep reinforcement learning (DRL) techniques for adjusting cell parameters of a wireless network by generating a plurality of experience tuples, and updating the neural network based on the generated experience tuples. The trained neural network may be used to select actions to adjust the cell parameters. Each experience tuple includes a cell identifier, a first state, a second state, an action applied to the cell that moves the cell from the first state to the second state, a local reward, and a global reward. The neural network is updated based on whether or not each action is acceptable, which is determined based on the global reward and the local reward associated with each action.

System and method for deep learning and wireless network optimization using deep learning

Embodiments of this disclosure filter a set of key performance indicators (KPIs) based on various criteria to reduce the number of root causes that are considered when adjusting the wireless configuration parameters. In particular, the set of KPIs may be filtered based on various criteria. In one example, KPIs having a hit-ratio below a hit-ratio threshold are removed from the set of KPIs. In another example, KPIs having a slope of linear regression below a threshold slope are removed from the set of KPIs. In yet another example, a KPI is removed from the set of KPIs when a causal relationship between the KPI and another KPI in the set of KPIs exceeds a threshold. In yet another example, a KPI is removed from the set of KPIs when the KPI has a prediction impact score that fails to exceed a prediction impact threshold.

Systems and methods for identifying causes of quality degradation in wireless networks

The present technology provides a new approach to optimizing wireless networks, including the coverage and capacity of cellular networks, using deep learning. The proposed method involves generating a group of cells comprising a cell identified as underperforming and one or more neighboring cells, ranking the one or more neighboring cells based on one or more relationship parameters between the underperforming cell and the one or more neighboring cells, and generating a multi-dimensional multi-channel state tensor for the group of cells based on the ranking of the one or more neighboring cells. This approach to cellular network optimization improves the coverage and capacity of cellular networks using a process that is faster, more accurate, less costly, and more robust.

Optimizing cellular networks using deep learning

Convergence times associated with simulated annealing based (SA-based) optimization in wireless networks can be reduced by introducing an additional local or cell-level evaluation step into the evaluation of global solutions. In particular, new local solutions may be evaluated based on local performance criteria when the new solutions are in a global solution deemed to have satisfied a global performance criteria. New local solutions that satisfy their corresponding local performance criteria remain in the new global solution. New local solutions that do not satisfy their corresponding local performance criteria are replaced with a corresponding current local solution from a current global solution, thereby modifying the new global solution. The resulting modified global solution includes both new local solutions and current local solutions prior to being accepted as the current global solution for the next iteration.

Autonomous, closed-loop and adaptive simulated annealing based machine learning approach for intelligent analytics-assisted self-organizing-networks (SONs)

Quality of service problems in wireless networks can be identified, and in some cases resolved, autonomously using an optimization technique that adjusts parameters of the wireless network based on relationships between key quality indicators (KQIs) and key performance indicators (KPIs). A KQI may be used to gauge the quality of service/experience collectively observed by users/devices when communicating a particular type of traffic in a wireless network or wireless network area. A KPI may be any specific performance metric of a wireless network tending to have a causal or correlative relationship with a KQI. Different types of relationships between a KQI and KPIs may be used to evaluate a quality of service problem, including correlation coefficients, slopes of linear regression, hit-ratios, and others.

Systems and methods for robustly determining time series relationships in wireless networks

The strengths of alternative self-organizing-network (SON) techniques can be leveraged by deriving a compromise result from alternative results generated by the respective SON techniques. In particular, the compromise result may be derived from the alternative results based on reputations assigned to alternative SON techniques used to generate the respective results. The compromise result may be calculated based on weighted averages of the alternative results (e.g., solutions, diagnoses, predicted values, etc.), or on weighted averages of parameters specified by the alternative results (e.g., parameter adjustments, underlying causes, KPI values, etc.). In such an embodiment, the weights applied to the alternative results may be based on the reputations of the corresponding SON techniques used to generate the respective alternative results.

Analytics-assisted, multi-agents, self-learning, self-managing, flexible and adaptive framework for intelligent SON

A system and method of predicting cellular network performance based on observed performance indicator data from cells in the cellular network. The method includes accessing a set of observed performance indicator data, the performance indicator data including a time sequenced measure of performance indicators for the cellular network. The method then classifies the observed performance data based on a cell as one of a high load growth cell and a high load non-growth cell. Based on the classification of the cell, the method computes a future value of at least one performance indicator using a predictive model based on testing data for the cell in the observed performance indicator data. The predictive model is derived from training data in the observed performance indicator data. An indication of the future value of the at least one of the performance indicators is output when the future value exceeds an alarm value.

Prediction of performance indicators in cellular networks

Adjusting cell specific antenna configuration parameters includes receiving, at each of a plurality of radio access nodes in a network, measurement reports from a plurality of user equipment devices. Base incremental adjustments to configuration parameters of one or more antennas are performed at each radio access node in response to the measurement reports. Additional measurement reports are received from the plurality of user equipment devices after the incremental adjustments. Base incremental adjustments to the configuration parameters of the one or more antennas at the radio access nodes continue to be performed in response to the measurement reports after previous incremental adjustments until an improvement limit has occurred. Biased random adjustments to the configuration parameters of the one or more antennas at the radio access nodes are performed in response to the improvement limit until a desired optimum result is achieved.

Method and apparatus for optimizing cell specific antenna configuration parameters

Methods and apparatus for detecting a traffic suppression turning point in a communications system based on traffic behavior are provided. Models representing relationship between traffic loads and a key performance indicator of a cell or a cluster of cells may be built and tested to generate a set of prediction errors corresponding to a plurality of traffic load ranges. The prediction errors are examined against a criteria to determine a traffic suppression turning point in terms of traffic loads. The models built may also be used to calculate a set of KPI slope values corresponding to different traffic load ranges. The set of KPI slope values are examined against a criteria to determine a traffic suppression turning point in terms of traffic loads.

Growing community of inventors

Bridgewater, NJ, United States of America

Jin Yang