The paper addresses the one-class classification (OCC) problem and advocates a one-class multiple kernel learning (MKL) approach for this purpose. To this aim, based on the Fisher null-space one-class classification principle, we present a multiple kernel learning algorithm where an $\ell_p$-norm constraint ($p\geq1$) on kernel weights is considered. We cast the proposed one-class MKL task as a min-max saddle point Lagrangian optimisation problem and propose an efficient method to solve it. An extension of the proposed one-class MKL approach is also considered where several related one-class MKL tasks are learned concurrently by constraining them to share common kernel weights. An extensive assessment of the proposed method on a range of data sets from different application domains confirms its merits against the baseline and several other algorithms.

Validated both numerically and experimentally, Beam Propagation Method (BPM) has been proven to be a very efficient and reasonably accurate simulation approach for certain silicon photonics (SiP) devices. This paper clears up some misunderstanding in SiP community that BPM is not suitable for SiP devices.

More than three hundred years ago: Newton thought that the reason of apple falling was due to gravity, although he discovered the gravity in the earth, he did not know what how produce gravitation; he said that God is the first driving force. From then on, more than three centuries have passed many people had been exploring the origin of gravitation, until today people have no notion on the origin of gravitation. In order to find the origin of gravitation, I did some experiments and discovered that moving photons generate gravitation. This discovery come to light the origin of gravitation: applying this discovery to others particles show the same effects; in the same time these discoveries have been applied in macro world to calculate the acceleration of gravity, first time display the more accurately formula than Universal law of gravitation to describe the gravity field of earth. These works indicate that moving elementary particles bring on interaction force which appearing behavior of gravitation in the macro world. This show that these discoveries reveal the essence of Universal law of gravitation from micro world to macro world, and reveal the origin of gravitation: namely, tell us where gravitation come from, and solve the enigmas what Universal law of gravitation cannot solve until today.

Continuous Integration and Continuous Delivery (CI/CD) is one of the widely discussed topics in the recent information technology (IT) world within the software engineering and delivery sectors. CI/CD is an Agile based DevOps methodology in software engineering that has proven to help improved quality and efficiency in the software delivery process which results in cost effective and low risk business solutions. With the growing demand in the global IT industry specifically in the DevOps based software engineering, estimated to reach $8 billion by 2022, up from $3.9 billion in 2017 [1], we commonly see that more focus is now being given to process optimization to attain efficiencies. Sri Lanka has always been one of the topmost sought-after destinations for outsourced IT labor in the world and currently employs 115, 000 skilled IT professionals [2] and it is important to include the advancements to process optimization to gain better results and keep up that momentum going. CI/CD being classified in the international IT market an improvement as such, the adoption of CI/CD and its perceived impact among Sri Lankan IT teams were researched. The literature review carried out did not provide much detail based on statistical evidence to understand the current adoption of CI/CD as well as the degree to which it is adopted among the IT teams specific to the Sri Lankan context. If CI/CD is adopted within Sri Lanka already, although undocumented, it was also important to understand what factors impact the adoption rate. Agile development methodologies are often adaptable to suit the best working solutions for teams, companies and even industries, hence there can be major inapplicability when it comes to studying a global set of factors from different contexts impacting the Sri Lankan teams. Building from this, the three top factors stated in international literature, team factors, stakeholder buy-in and complexity were studied, specifically within the Sri Lankan context to evaluate their impact on CI/CD adoption.

Prediction of future movement of stock prices has been a subject matter of many research work. On one hand, we have proponents of the Efficient Market Hypothesis who claim that stock prices cannot be predicted, on the other hand, there are propositions illustrating that, if appropriately modeled, stock prices can be predicted with a high level of accuracy. There is also a gamut of literature on technical analysis of stock prices where the objective is to identify patterns in stock price movements and profit from it. In this work, we propose a hybrid approach for stock price prediction using five deep learning-based regression models. We select the NIFTY 50 index values of the National Stock Exchange (NSE) of India, over a period of December 29, 2014 to July 31, 2020. Based on the NIFTY data during December 29, 2014 to December 28, 2018, we build two regression models using convolutional neural networks (CNNs), and three regression models using long-and-short-term memory (LSTM) networks for predicting the open values of the NIFTY 50 index records for the period December 31, 2018 to July 31, 2020. We adopted a multi-step prediction technique with walk-forward validation. The parameters of the five deep learning models are optimized using the grid-search technique so that the validation losses of the models stabilize with an increasing number of epochs in the model training, and the training and validation accuracies converge. Extensive results are presented on various metrics for all the proposed regression models. The results indicate that while both CNN and LSTM-based regression models are very accurate in forecasting the NIFTY 50 open values, the CNN model that previous one week’s data as the input is the fastest in its execution. On the other hand, the encoder-decoder convolutional LSTM model uses the previous two weeks’ data as the input is found to be the most accurate in its forecasting results.

In many scientific and engineering applications, Bayesian optimization (BO) is a powerful tool for hyperparameter tuning of a machine learning model, materials design and discovery, etc. BO guides the choice of experiments in a sequential way to find a good combination of design points in as few experiments as possible. It can be formulated as a problem of optimizing a “black-box” function. Different from single-task Bayesian optimization, Multi-task Bayesian optimization is a general method to efficiently optimize multiple different but correlated “black-box” functions. The previous works in Multi-task Bayesian optimization algorithm queries a point to be evaluated for all tasks in each round of search, which is not efficient. For the case where different tasks are correlated, it is not necessary to evaluate all tasks for a given query point. Therefore, the objective of this work is to develop an algorithm for multi-task Bayesian optimization with automatic task selection so that only one task evaluation is needed per query round. Specifically, a new algorithm, namely, multi-task Gaussian process upper confidence bound (MT-GPUCB), is proposed to achieve this objective. The MT-GPUCB is a two-step algorithm, where the first step chooses which query point to evaluate, and the second step automatically selects the most informative task to evaluate. Under the bandit setting, a theoretical analysis is provided to show that our proposed MT-GPUCB is no-regret under some mild conditions. Our proposed algorithm is verified experimentally on a range of synthetic functions as well as real-world problems. The results clearly show the advantages of our query strategy for both design point and task.

Abstract—Non-orthogonal multiplexing (NOM) is a novel superposition coding inspired scheme that has been recently proposed for improving the power, spectrum efficiency and delay of wireless links with packet error rate (PER) constraints. Despite its efficiency, restricting the number of multiplexed packets to two limits the throughput improvement to 100%. Therefore, this work presents a novel NOM design with unlimited number of multiplexed packets by manipulating the repeated transmissions in automatic repeat request (ARQ) to enhance the power and spectrum efficiency by multiplexing new and repeated packets while taking into account the channel conditions and varying the power per packet in different transmissions. The proposed scheme employs an efficient heuristic algorithm to perform the power assignment and multiplexing decisions. Moreover, the complexity of the proposed NOM can be controlled by enforcing a limit on the maximum number of multiplexed packets per transmission, making it suitable for different types of Internet of Things (IoT) nodes with various computational capabilities. The obtained results demonstrate the effectiveness of proposed scheme, which offers up to 200% spectral efficiency improvement at moderate signal to noise ratios (SNRs), and up to 700% at high SNRs. Furthermore, the new scheme can reduce the transmission power consumption by up to 6 dB in the high SNR region.

This paper presented an regrasp planning method to eliminate grasp uncertainty while considering the geometric constraints of a fixture. The method automatically finds the Stable Placement Poses (SPPs) of an object on a Triangular Corner Fixture (TCF), elevates the object from its SPPs to dropping poses and finds the Deterministic Dropping Poses (DDPs), builds regrasp graphs by using the SPP-DDP pairs and their associated grasp configurations, and searches the graph to find regrasp motion sequences for precise assembly. Since the SPPs and their associated regrasps are constrained by the TCF’s geometry and have high precision, the final object poses regrasped via it has low uncertainty and can be directly used for assembly by position control. In the experimental section, we study the performance of analytical and learning-based methods for estimating the DDPs of different objects and quantitatively examine the proposed method’s ability to suppress uncertainty using assembly tasks like peg-in-hole insertion and sheathing tubes, aligning holes, mounting bearing housings, etc. The results demonstrate the method’s robustness and efficacy.

We have improved an article called ‘Multi-key FHE from LWE, revisited’ in TCC’16 and proposed a Dynamic multi-key FHE in asymmetric key setting from LWE. Can you give me some suggestions for modification? thanks very much!

In competitive electricity markets the optimal trading problem of an electricity market agent is commonly formulated as a bi-level program, and solved as mathematical program with equilibrium constraints (MPEC). In this paper, an alternative paradigm, labeled as mathematical program with neural network constraint (MPNNC), is developed to incorporate complex market dynamics in the optimal bidding strategy. This method uses input-convex neural networks (ICNNs) to represent the mapping between the upper-level (agent) decisions and the lower-level (market) outcomes, i.e., to replace the lower-level problem by a neural network. In a comparative analysis, the optimal bidding problem of a load agent is formulated via the proposed MPNNC and via the classical bi-level programming method, and compared against each other.

Prediction of stock prices using time series analysis is quite a difficult and challenging task since the stock prices usually depict random patterns of movement. However, the last decade has witnessed rapid development and evolution of sophisticated algorithms for complex statistical analysis. These algorithms are capable of processing a large volume of time series data executing on high-performance hardware and parallel computing architecture. Thus computations which were seemingly impossible to perform a few years back are quite amenable to real-time time processing and effective analysis today. Stock market time series data are large in volume, and quite often need real-time processing and analysis. Thus it is quite natural that research community has focused on designing and developing robust predictive models for accurately forecasting stochastic nature of stock price movements. This work presents a time series decomposition-based approach for understanding the past behavior of the realty sector of India, and forecasting its behavior in future. While the forecasting models are built using the time series data of the realty sector for the period January 2010 till December 2015, the prediction is made for the time series index values for the months of the year 2016. A detailed comparative analysis of the methods are presented with respect to their forecasting accuracy and extensive results are provided to demonstrate the effectiveness of the six proposed forecasting models.

In recent years, DNA-based systems have become a promising medium for long-term data storage. There are two layers of errors in DNA-based storage systems. The first is the dropouts of the DNA strands, which has been characterized in the shuffling-sampling channel. The second is insertions, deletions, and substitutions of nucleotides in individual DNA molecules. In this paper, we describe a DNA noisy synchronization error channel to characterize the errors in individual DNA molecules. We derive non-trivial lower and upper capacity bounds of the DNA noisy synchronization error channel based on information theory. By cascading these two channels, we provide theoretical capacity limits of the DNA storage system. These results reaffirm that DNA is a reliable storage medium with high storage density potential.

Time series analysis and forecasting of stock market prices has been a very active area of research over the last two decades. Availability of extremely fast and parallel architecture of computing and sophisticated algorithms has made it possible to extract, store, process and analyze high volume stock market time series data very efficiently. In this paper, we have used time series data of the two sectors of the Indian economy – Information Technology (IT) and Capital Goods (CG) for the period January 2009 – April 2016 and have studied the relationships of these two time series with the time series of DJIA indices, NIFTY indices and the US Dollar to Indian Rupees exchange rate. We established by graphical and statistical tests that while the IT sector of India has a strong association with DJIA indices and the Dollar to Rupee exchange rate, the Indian CG sector exhibits a strong association with the NIFTY indices. We contend that these observations corroborate our hypotheses that the Indian IT sector is strongly coupled with the world economy whereas the CG sector of India is the reflection of India’s internal economic growth. We also present several models of regression between the time series which exhibit strong association among them. The effectiveness of these models have been demonstrated by very low values of their forecasting errors.

Designing efficient and robust algorithms for accurate prediction of stock market prices is one of the most exciting challenges in the field of time series analysis and forecasting. With the exponential rate of development and evolution of sophisticated algorithms and with the availability of fast computing platforms, it has now become possible to effectively and efficiently extract, store, process and analyze high volume of stock market data with diversity in its contents. Availability of complex algorithms which can execute very fast on parallel architecture over the cloud has made it possible to achieve higher accuracy in forecasting results while reducing the time required for computation. In this paper, we use the time series data of the healthcare sector of India for the period January 2010 till December 2016. We first demonstrate a decomposition approach of the time series and then illustrate how the decomposition results provide us with useful insights into the behavior and properties exhibited by the time series. Further, based on the structural analysis of the time series, we propose six different methods of forecasting for predicting the time series index of the healthcare sector. Extensive results are provided on the performance of the forecasting methods to demonstrate their effectiveness.

This letter considers minimizing the bit error rate (BER) of unmanned aerial vehicle (UAV) communications assisted by intelligent reflecting surfaces (IRSs). By noting that increasing the number of IRS elements in the presence of phase errors does not necessarily improve the BER, it is crucial to use only the elements that contribute to reducing the BER. Consequently, we propose an efficient algorithm to activate only the elements that improve the BER. The proposed algorithm has lower complexity and comparable BER to the optimum selection process, which is an NP-hard problem. The accuracy of the estimated phase is evaluated by deriving the probability distribution function (PDF) of the least-square (LS) channel estimator, and showing that the PDF can be closely approximated by the von Mises distribution at high signal-to-noise ratios (SNRs). The obtained analytical and simulation results show that using all the available reflectors can significantly deteriorate the BER, and thus, elements’ selection is necessary. In particular scenarios, using about 26% of the reflectors provides more than 10 fold BER reduction.

A document by Sertse Abebe Ayalew . Click on the document to view its contents.

We propose and analyze a compact and non-volatile nanomagnetic (all-spin) analog matrix multiplier performing the multiply-and-accumulate (MAC) operation using two magnetic tunnel junctions – one activated by strain to act as the multiplier, and the other activated by spin-orbit torque pulses to act as a domain wall synapse that performs the operation of the accumulator. Each MAC operation can be performed in ~1 ns and the maximum energy dissipated per operation is ~100 aJ. This provides a very useful hardware accelerator for machine learning (e.g. training of deep neural networks), solving combinatorial optimization problems with Ising type machines, and other artificial intelligence tasks which often involve the multiplication of large matrices. The non-volatility allows the matrix multiplier to be embedded in powerful non-von-Neumann architectures.

In this paper we consider such HAPS-based cellular systems to provide IoT coverage for protected area.