Umair Ali Khan

Online learning of timeout policies for dynamic power management

Dynamic power management (DPM) refers to strategies which selectively change the operational states of a device during runtime to reduce the power consumption based on the past usage pattern, the current workload, and the given performance constraint. The power management problem becomes more challenging when the workload exhibits nonstationary behavior which may degrade the performance of any single or static DPM policy. This article presents a reinforcement learning (RL)-based DPM technique for optimal selection of timeout values in the different device states. Each timeout period determines how long the device will remain in a particular state before the transition decision is taken. The timeout selection is based on workload estimates derived from a Multilayer Artificial Neural Network (ML-ANN) and an objective function given by weighted performance and power parameters. Our DPM approach is further able to adapt the power-performance weights online to meet user-specified power and performance constraints, respectively. We have completely implemented our DPM algorithm on our embedded traffic surveillance platform and performed long-term experiments using real traffic data to demonstrate the effectiveness of the DPM. Our results show that the proposed learning algorithm not only adequately explores the power-performance trade-off with nonstationary workload but can also successfully perform online adjustment of the trade-off parameter in order to meet the user-specified constraint.

A Reinforcement Learning Framework for Dynamic Power Management of a Portable, Multi-camera Traffic Monitoring System

Dynamic Power Management (DPM) refers to a set of strategies that achieves efficient power consumption by selectively turning off (or reducing the performance of) a system components when they are idle or are serving light workloads. This paper presents a Reinforcement Learning (RL) based DPM technique for a portable, multi-camera traffic monitoring system. We target the computing hardware of the sensing platform which is the major contributor to the entire power consumption. The RL technique used for the DPM of the sensing platform uses a model-free learning algorithm that does not require a priori model of the system. In addition, a robust workload estimator based on an online, Multi-Layer Artificial Neural Network (ML-ANN) is incorporated to the learning algorithm to provide partial information about the workload and to take better decisions according to the changing workload. Based on the estimated workload and a selected power-latency tradeoff parameter, the algorithm learns to use optimal time-out values in sleep and idle modes of the computing hardware. Our results show that the learning algorithm learns an optimal DPM policy for the non-stationary workload, while significantly reducing the power consumption and keeping the system response to a desired level.

Performance Evaluation of Mobile Ad Hoc Routing Mechanisms

Abstract Routing protocols play a pivotal role in energy-efficient, reliable and robust communication in Mobile Ad hoc Networks (MANETs). In order to ensure efficient communication, the optimal operation setting of a routing protocol is essential to be ascertained. In this paper, we perform a comparative analysis of three most popular MANET routing protocols, namely, ad hoc on demand distance vector (AODV), dynamic source routing, and destination sequenced distance vector protocols. We evaluate the performance of these protocols for different network sizes, each with low and high traffic scenario. The generic evaluation criteria which specify the performance of routing protocols and used in our simulations include packet delivery ratio, end-to-end delay, average remaining energy of nodes, and throughput. Our in-depth analysis and the comparison results presented in this paper show that AODV protocol outperforms the other two protocols for the selected parameters and various network scenarios.

Deep learning based beat event detection in action movie franchises

Automatic understanding and interpretation of movies can be used in a variety of ways to semantically manage the massive volumes of movies data. “Action Movie Franchises” dataset is a collection of twenty Hollywood action movies from five famous franchises with ground truth annotations at shot and beat level of each movie. In this dataset, the annotations are provided for eleven semantic beat categories. In this work, we propose a deep learning based method to classify shots and beat-events on this dataset. The training dataset for each of the eleven beat categories is developed and then a Convolution Neural Network is trained. After finding the shot boundaries, key frames are extracted for each shot and then three classification labels are assigned to each key frame. The classification labels for each of the key frames in a particular shot are then used to assign a unique label to each shot. A simple sliding window based method is then used to group adjacent shots having the same label in order to find a particular beat event. The results of beat event classification are presented based on criteria of precision, recall, and F-measure. The results are compared with the existing technique and significant improvements are recorded.

I-RP: Interference Aware Routing Protocol for WBAN.

The Wireless Body Sensor Networks (WBSN) have witnessed tremendous research interest because of their wide range of applications (medical and non-medical) in order to improve the quality of life. The healthcare applications of WBSN demands dissemination of patient’s data, reliably and in a timely manner. For this purpose, medical teams may use real-time applications for disseminating critical data such as blood pressure, ECG, and EEG. The critical data packets are highly delay sensitive that must reach intended destination within time constraints. Due to the exchange of real-time and multi-media data, some nodes or links may experience the significant level of interference in the network. Consequently, it results in transmission disruption, random number of packet drops, insufficient buffer space and lack of availability of bandwidth. Moreover, interference in the network strains the communication links, reduces the information delivery capacity of the network and leads to high collisions, packet losses, retransmission and energy consumption. Therefore, incorporating interference-awareness in routing decisions is desirable to enhance the performance of WBSN. In this paper, we present an Interference-aware Routing Protocol (I-RP) that makes use of composite routing metric incorporating link quality (in terms of link delay and interference level) and path length. This multi-facet routing strategy makes more informed routing decision regarding route selection in a way that, a route with the minimum level of interference and path length is selected. Moreover, it also increases the link reliability and minimizes the packet losses and retransmission. The simulation results demonstrate the improved performance of proposed scheme when compared to existing routing scheme in WBSN.

Movies tags extraction using deep learning

Retrieving information from movies is becoming increasingly demanding due to the enormous amount of multimedia data generated each day. Not only it helps in efficient search, archiving and classification of movies, but is also instrumental in content censorship and recommendation systems. Extracting key information from a movie and summarizing it in a few tags which best describe the movie presents a dedicated challenge and requires an intelligent approach to automatically analyze the movie. In this paper, we formulate movies tags extraction problem as a machine learning classification problem and train a Convolution Neural Network (CNN) on a carefully constructed tag vocabulary. Our proposed technique first extracts key frames from a movie and applies the trained classifier on the key frames. The predictions from the classifier are assigned scores and are filtered based on their relative strengths to generate a compact set of most relevant key tags. We performed a rigorous subjective evaluation of our proposed technique for a wide variety of movies with different experiments. The evaluation results presented in this paper demonstrate that our proposed approach can efficiently extract the key tags of a movie with a good accuracy.

Machine Learning based Number Plate Detection and Recognition

Automatic Number Plate Detection and Recognition (ANPDR) has become of significant interest with the substantial increase in the number of vehicles all over the world. ANPDR is particularly important for automatic toll collection, traffic law enforcement, parking lot access control, and gate entry control, etc. Due to the known efficacy of image processing in this context, a number of ANPDR solutions have been proposed. However, these solutions are either limited in operations or work only under specific conditions and environments. In this paper, we propose a robust and computationally-efficient ANPDR system which uses Deformable Part Models (DPM) for extracting number plate features from training images, Structural Support Vector Machine (SSVM) for training a number plate detector with the extracted DPM features, several image enhancement operations on the extracted number plate, and Optical Character Recognition (OCR) for extracting the numbers from the plate. The results presented in this paper, obtained by long-term experiments performed under different conditions, demonstrate the efficiency of our system. They also show that our proposed system outperforms other ANPDR techniques not only in accuracy, but also in execution time.