Double Hybrid Functionals and the Π-System Bond Length Alternation Challenge: Rivaling Accuracy of Post-HF Methods

Predicting accurate bond length alternations (BLAs) in long conjugated oligomers has been a significant challenge for electronic-structure methods for many decades, made particularly important by the close relationships between BLA and the rich optoelectronic properties of π-delocalized systems. Here, we test the accuracy of recently developed, and increasingly popular, double hybrid (DH) functionals, positioned at the top of Jacobs Ladder of DFT methods of increasing sophistication, computational cost, and accuracy, due to incorporation of MP2 correlation energy. Our test systems comprise oligomeric series of polyacetylene, polymethineimine, and polysilaacetylene up to six units long. MP2 calculations reveal a pronounced shift in BLAs between the 6-31G(d) basis set used in many studies of BLA to date and the larger cc-pVTZ basis set, but only modest shifts between cc-pVTZ and aug-cc-pVQZ results. We hence perform new reference CCSD(T)/cc-pVTZ calculations for all three series of oligomers against which we assess the performance of several families of DH functionals based on BLYP, PBE, and TPSS, along with lower-rung relatives including global- and range-separated hybrids. Our results show that DH functionals systematically improve the accuracy of BLAs relative to single hybrid functionals. xDH-PBE0 (N4 scaling using SOS-MP2) emerges as a DH functional rivaling the BLA accuracy of SCS-MP2 (N5 scaling), which was found to offer the best compromise between computational cost and accuracy the last time the BLA accuracy of DFT- and wave function-based methods was systematically investigated. Interestingly, xDH-PBE0 (XYG3), which differs to other DHs in that its MP2 term uses PBE0 (B3LYP) orbitals that are not self-consistent with the DH functional, is an outlier of trends of decreasing average BLA errors with increasing fractions of MP2 correlation and HF exchange.

Accurate Treatment of Large Supramolecular Complexes by Double-Hybrid Density Functionals Coupled with Nonlocal van der Waals Corrections

In this work, we present a thorough assessment of the performance of some representative double-hybrid density functionals (revPBE0-DH-NL and B2PLYP-NL) as well as their parent hybrid and GGA counterparts, in combination with the most modern version of the nonlocal (NL) van der Waals correction to describe very large weakly interacting molecular systems dominated by noncovalent interactions. Prior to the assessment, an accurate and homogeneous set of reference interaction energies was computed for the supramolecular complexes constituting the L7 and S12L data sets by using the novel, precise, and efficient DLPNO-CCSD(T) method at the complete basis set limit (CBS). The correction of the basis set superposition error and the inclusion of the deformation energies (for the S12L set) have been crucial for obtaining precise DLPNO-CCSD(T)/CBS interaction energies. Among the density functionals evaluated, the double-hybrid revPBE0-DH-NL and B2PLYP-NL with the three-body dispersion correction provide remarkably accurate association energies very close to the chemical accuracy. Overall, the NL van der Waals approach combined with proper density functionals can be seen as an accurate and affordable computational tool for the modeling of large weakly bonded supramolecular systems.

Modern Computational Chemistry Methods for Prediction of Ground- and Excited-State Properties in Open-Shell Systems

The main goal of the research presented in this work is to provide some important insights about computational modeling of open-shell species. Such projects are: the investigation of the size-extensivity error in Equation-of-Motion Coupled Cluster methods, the analysis of the Long-Range corrected scheme in predicting UV-Vis spectra of Cu(II) complexes with the 4-imidazole acetate and its ethylated derivative, and the exploration of the importance of choosing a proper basis set for the description of systems such as the lithium monoxide anion. The most significant findings of this research are: (i) The contribution of the left operator to the size-extensivity error of the CR-EOMCC(2,3) approach, (ii) The cause of d-d shifts when varying the range-separation parameter and the amount of the exact exchange arising from the imbalanced treatment of localized vs. delocalized orbitals via the "tuned" CAM-B3LYP* functional, (iii) The proper acidity trend of the first-row hydrides and their lithiated analogs that may be reversed if the basis sets are not correctly selected.

Image analysis by moment invariants using a set of step-like basis functions

Moment invariants have been thoroughly studied and repeatedly proposed as one of the most powerful tools for 2D shape identification. In this paper a set of such descriptors is proposed, being the basis functions discontinuous in a finite number of points. The goal of using discontinuous functions is to avoid the Gibbs phenomenon, and therefore to yield a better approximation capability for discontinuous signals, as images. Moreover, the proposed set of moments allows the definition of rotation invariants, being this the other main design concern. Translation and scale invariance are achieved by means of standard image normalization. Tests are conducted to evaluate the behavior of these descriptors in noisy environments, where images are corrupted with Gaussian noise up to different SNR values. Results are compared to those obtained using Zernike moments, showing that the proposed descriptor has the same performance in image retrieval tasks in noisy environments, but demanding much less computational power for every stage in the query chain.

Validation-Based Sparse Gaussian Process Classifier Design

Gaussian processes (GPs) are promising Bayesian methods for classification and regression problems. Design of a GP classifier and making predictions using it is, however, computationally demanding, especially when the training set size is large. Sparse GP classifiers are known to overcome this limitation. In this letter, we propose and study a validation-based method for sparse GP classifier design. The proposed method uses a negative log predictive (NLP) loss measure, which is easy to compute for GP models. We use this measure for both basis vector selection and hyperparameter adaptation. The experimental results on several real-world benchmark data sets show better orcomparable generalization performance over existing methods.

Finite Signal-set Capacity of Two-user Gaussian Multiple Access Channel

The capacity region of a two-user Gaussian Multiple Access Channel (GMAC) with complex finite input alphabets and continuous output alphabet is studied. When both the users are equipped with the same code alphabet, it is shown that, rotation of one of the user’s alphabets by an appropriate angle can make the new pair of alphabets not only uniquely decodable, but will result in enlargement of the capacity region. For this set-up, we identify the primary problem to be finding appropriate angle(s) of rotation between the alphabets such that the capacity region is maximally enlarged. It is shown that the angle of rotation which provides maximum enlargement of the capacity region also minimizes the union bound on the probability of error of the sumalphabet and vice-verse. The optimum angle(s) of rotation varies with the SNR. Through simulations, optimal angle(s) of rotation that gives maximum enlargement of the capacity region of GMAC with some well known alphabets such as M-QAM and M-PSK for some M are presented for several values of SNR. It is shown that for large number of points in the alphabets, capacity gains due to rotations progressively reduce. As the number of points N tends to infinity, our results match the results in the literature wherein the capacity region of the Gaussian code alphabet doesn’t change with rotation for any SNR.

Structured precision modelling with Cholesky basis superposition for speech recognition

Structured precision modelling is an important approach to improve the intra-frame correlation modelling of the standard HMM, where Gaussian mixture model with diagonal covariance are used. Previous work has all been focused on direct structured representation of the precision matrices. In this paper, a new framework is proposed, where the structure of the Cholesky square root of the precision matrix is investigated, referred to as Cholesky Basis Superposition (CBS). Each Cholesky matrix associated with a particular Gaussian distribution is represented as a linear combination of a set of Gaussian independent basis upper-triangular matrices. Efficient optimization methods are derived for both combination weights and basis matrices. Experiments on a Chinese dictation task showed that the proposed approach can significantly outperformed the direct structured precision modelling with similar number of parameters as well as full covariance modelling. © 2011 IEEE.

A novel radial basis function neural network principal component analysis scheme for PMU-based wide-area power system monitoring

A novel model-based principal component analysis (PCA) method is proposed in this paper for wide-area power system monitoring, aiming to tackle one of the critical drawbacks of the conventional PCA, i.e. the incapability to handle non-Gaussian distributed variables. It is a significant extension of the original PCA method which has already shown to outperform traditional methods like rate-of-change-of-frequency (ROCOF). The ROCOF method is quick for processing local information, but its threshold is difficult to determine and nuisance tripping may easily occur. The proposed model-based PCA method uses a radial basis function neural network (RBFNN) model to handle the nonlinearity in the data set to solve the no-Gaussian issue, before the PCA method is used for islanding detection. To build an effective RBFNN model, this paper first uses a fast input selection method to remove insignificant neural inputs. Next, a heuristic optimization technique namely Teaching-Learning-Based-Optimization (TLBO) is adopted to tune the nonlinear parameters in the RBF neurons to build the optimized model. The novel RBFNN based PCA monitoring scheme is then employed for wide-area monitoring using the residuals between the model outputs and the real PMU measurements. Experimental results confirm the efficiency and effectiveness of the proposed method in monitoring a suite of process variables with different distribution characteristics, showing that the proposed RBFNN PCA method is a reliable scheme as an effective extension to the linear PCA method.

Comparing Support Vector Machines with Gaussian Kernels to Radial Basis Function Classifiers

The Support Vector (SV) machine is a novel type of learning machine, based on statistical learning theory, which contains polynomial classifiers, neural networks, and radial basis function (RBF) networks as special cases. In the RBF case, the SV algorithm automatically determines centers, weights and threshold such as to minimize an upper bound on the expected test error. The present study is devoted to an experimental comparison of these machines with a classical approach, where the centers are determined by $k$--means clustering and the weights are found using error backpropagation. We consider three machines, namely a classical RBF machine, an SV machine with Gaussian kernel, and a hybrid system with the centers determined by the SV method and the weights trained by error backpropagation. Our results show that on the US postal service database of handwritten digits, the SV machine achieves the highest test accuracy, followed by the hybrid approach. The SV approach is thus not only theoretically well--founded, but also superior in a practical application.

Robust initialisation of Gaussian radial basis function networks using partitioned k-means clustering

Radial basis function networks can be trained quickly using linear optimisation once centres and other associated parameters have been initialised. The authors propose a small adjustment to a well accepted initialisation algorithm which improves the network accuracy over a range of problems. The algorithm is described and results are presented.

Genetic algorithms and Gaussian Bayesian networks to uncover the predictive core set of bibliometric indices

The diversity of bibliometric indices today poses the challenge of exploiting the relationships among them. Our research uncovers the best core set of relevant indices for predicting other bibliometric indices. An added difficulty is to select the role of each variable, that is, which bibliometric indices are predictive variables and which are response variables. This results in a novel multioutput regression problem where the role of each variable (predictor or response) is unknown beforehand. We use Gaussian Bayesian networks to solve the this problem and discover multivariate relationships among bibliometric indices. These networks are learnt by a genetic algorithm that looks for the optimal models that best predict bibliometric data. Results show that the optimal induced Gaussian Bayesian networks corroborate previous relationships between several indices, but also suggest new, previously unreported interactions. An extended analysis of the best model illustrates that a set of 12 bibliometric indices can be accurately predicted using only a smaller predictive core subset composed of citations, g-index, q2-index, and hr-index. This research is performed using bibliometric data on Spanish full professors associated with the computer science area.

Gaussian processes:iterative sparse approximations

In recent years there has been an increased interest in applying non-parametric methods to real-world problems. Significant research has been devoted to Gaussian processes (GPs) due to their increased flexibility when compared with parametric models. These methods use Bayesian learning, which generally leads to analytically intractable posteriors. This thesis proposes a two-step solution to construct a probabilistic approximation to the posterior. In the first step we adapt the Bayesian online learning to GPs: the final approximation to the posterior is the result of propagating the first and second moments of intermediate posteriors obtained by combining a new example with the previous approximation. The propagation of em functional forms is solved by showing the existence of a parametrisation to posterior moments that uses combinations of the kernel function at the training points, transforming the Bayesian online learning of functions into a parametric formulation. The drawback is the prohibitive quadratic scaling of the number of parameters with the size of the data, making the method inapplicable to large datasets. The second step solves the problem of the exploding parameter size and makes GPs applicable to arbitrarily large datasets. The approximation is based on a measure of distance between two GPs, the KL-divergence between GPs. This second approximation is with a constrained GP in which only a small subset of the whole training dataset is used to represent the GP. This subset is called the em Basis Vector, or BV set and the resulting GP is a sparse approximation to the true posterior. As this sparsity is based on the KL-minimisation, it is probabilistic and independent of the way the posterior approximation from the first step is obtained. We combine the sparse approximation with an extension to the Bayesian online algorithm that allows multiple iterations for each input and thus approximating a batch solution. The resulting sparse learning algorithm is a generic one: for different problems we only change the likelihood. The algorithm is applied to a variety of problems and we examine its performance both on more classical regression and classification tasks and to the data-assimilation and a simple density estimation problems.

Fuzzy Set Theory Approach as the Basis of Analysis of Financial Flows in the Economical Security System

The problems of formalization of the process of matching different management subjects’ functioning characteristics obtained on the financial flows analysis basis is considered. Formal generalizations for gaining economical security system knowledge bases elements are presented. One of feedback directions establishment between knowledge base of the system of economical security and financial flows database analysis is substantiated.

Best basis-based intelligent machine fault diagnosis

The wavelet packet transform decomposes a signal into a set of bases for time–frequency analysis. This decomposition creates an opportunity for implementing distributed data mining where features are extracted from different wavelet packet bases and served as feature vectors for applications. This paper presents a novel approach for integrated machine fault diagnosis based on localised wavelet packet bases of vibration signals. The best basis is firstly determined according to its classification capability. Data mining is then applied to extract features and local decisions are drawn using Bayesian inference. A final conclusion is reached using a weighted average method in data fusion. A case study on rolling element bearing diagnosis shows that this approach can greatly improve the accuracy ofdiagno sis.