Anila Usman

Wind power prediction using deep neural network based meta regression and transfer learning

Abstract An innovative short term wind power prediction system is proposed which exploits the learning ability of deep neural network based ensemble technique and the concept of transfer learning. In the proposed DNN-MRT scheme, deep auto-encoders act as base-regressors, whereas Deep Belief Network is used as a meta-regressor. Employing the concept of ensemble learning facilitates robust and collective decision on test data, whereas deep base and meta-regressors ultimately enhance the performance of the proposed DNN-MRT approach. The concept of transfer learning not only saves time required during training of a base-regressor on each individual wind farm dataset from scratch but also stipulates good weight initialization points for each of the wind farm for training. The effectiveness of the proposed, DNN-MRT technique is expressed by comparing statistical performance measures in terms of root mean squared error (RMSE), mean absolute error (MAE), and standard deviation error (SDE) with other existing techniques.

Review of Storage Techniques for Sparse Matrices

This paper reviews the current state of knowledge of the storage formats for sparse linear systems. Here we consider the ways developed so far for storing a sparse matrix and their quoted effects on computational speed. The main idea behind these formats involves keeping both the indices and the non-zero elements in the sparse matrix in a single data structure. These specialized schemes not only save storage but also yield computational savings. Since the locations of the non-zero elements in the matrix are known explicitly, unnecessary computations involving zeros can be avoided. Thus the use of these formats reduces additional memory required in the usual indexing based storage schemes and gives promising performance improvements

Implementation and Evaluation of Parallel Sparse Matrix-Vector Products on Distributed Memory Parallel Computers

The sparse matrix vector product (SMVP) is the kernel operation in many scientific applications. This kernel is an irregular problem, which has led to the development of several compressed storage formats. This paper discusses scalable implementations of sparse matrix-vector products, which are crucial for high performance solutions of large-scale linear equations, on distributed memory parallel computers using message passing. Five storage formats for sparse matrices are evaluated. We conduct numerical experiments on several different sparse matrices and show the parallel performance of our sparse matrix-vector product routines

Performance evaluation of storage formats for sparse matrices in fortran

Many storage formats have been proposed to represent spa- rse matrices. This paper extends to Fortran 95 the performance evaluation of sparse storage formats in Java presented at ICCS 2005, [7]. These experiments have the same set up (almost 200 sparse matrices and matrix-vector multiplication), but now consider the Fortran 95 Sparse BLAS reference implementation.

Storage formats for sparse matrices in java

Many storage formats (or data structures) have been proposed to represent sparse matrices. This paper presents a performance evaluation in Java comparing eight of the most popular formats plus one recently proposed specifically for Java (by Gundersen and Steihaug [6] – Java Sparse Array) using the matrix-vector multiplication operation.

Equilibrium states for predictor-corrector methods

A theory is developed that explains the stepsize patterns observed when standard predictor-corrector methods with variable stepsize strategy are used to solve stiff or mildly stiff problems. In some cases an algorithmic steady state occurs with smooth almost constant stepsizes; at other times an oscillating stepsize pattern of stepsizes is observed with the possibility of frequent rejected steps.

Alternative stepsize strategies for Adams predictor-corrector codes

Abstract Adams predictor–corrector methods are among the most widely used algorithms for solving initial value problems in ordinary differential equations. Adaptive stepsize techniques are employed to enhance the numerical stability and accuracy of these methods. This paper deals with the stepsize-control (SC) stability of Adams methods. The SC-stability conditions have been obtained for low-order Adams predictor–corrector methods using the standard stepsize strategies. However, when the stepsize is restricted by stability, it oscillates and frequent step rejections are observed. For the physically important case of real dominant eigenvalue of the Jacobian, the Adams methods are not SC-stable in general. In this paper, we investigate alternative stepsize strategies to smooth out the stepsize behaviour. It has been found that PI stepsize controller and estimation techniques, which have been developed for Runge–Kutta methods, fail to give good results in the case of Adams methods. A combined strategy has been formulated which eliminates stepsize oscillations and results in a smooth stepsize behaviour. All programming has been carried out in the integrated environment of standard software package MATLAB

Modified order and stepsize strategies in Adams codes

Abstract We derive some modifications to the order and stepsize strategies of an Adams multistep code. The objective is to produce smoother and more efficient handling of mildly stiff problems without affecting the codes performance on non-stiff problems. Results are given for the Van der Pol equation (before and after) to show the effect of the changes. We also consider introducing a modified predictor–corrector method, with an enlarged stability region, at low order.

Development of database auditing infrastructure

A database is considered as a core asset of an organization. Information about customers, employees and employers of the organization is stored in the database. Allowing authorized access to information and its shielding against unauthorized access is always given prime importance and is of great concern to the stakeholders of the organization. The work presented here outlines the development of a Database Auditing Infrastructure (DAI). The test bed for this infrastructure is chosen as MS SQL Server 2005. Database security is a vast and complicated discipline. Database auditing is the facilitating activity for implementing database security. Triggers, default trace, service brokers, queues, information from system databases are used in the implementation of DAI. Here a customized methodology, comprising of several ways has been adapted for implementing auditing in a database. The DAI presented in this paper is a generic one that can be easily integrated with an existing database. Once deployed the DAI starts providing auditing services to Information System Auditors (ISA) through its web based interface.

Steady state heat transfer using Galerkin finite element method

The finite-element methods (FEMs) have become one of the major strategies for solving partial differential equations. Galerkin method is one of widely used FEM technique to solve the problems in heat transfer, fluid mechanics and mechanical systems. The objective of this paper is to perform error analysis for different types of heat flow vectors using Galerkin method for triangular elements. The heat flow vectors applied in this article include constant, exponential, sinusoidal and polynomial functions. On the basis of simulation results some guidelines has been suggested for each category of these heat flow vector focusing on method accuracy.