Using non-parametric search algorithms to forecast daily excess stock returns

Are the learning procedures of genetic algorithms (GAs) able to generate optimal architectures for artificial neural networks (ANNs) in high frequency data? In this experimental study,GAs are used to identify the best architecture for ANNs. Additional learning is undertaken by the ANNs to forecast daily excess stock returns. No ANN architectures were able to outperform a random walk,despite the finding of non-linearity in the excess returns. This failure is attributed to the absence of suitable ANN structures and further implies that researchers need to be cautious when making inferences from ANN results that use high frequency data.

Forecasting exchange rates with linear and nonlinear models

In this paper the exchange rate forecasting performance of neural network models are evaluated against random walk and a range of time series models. There are no guidelines available that can be used to choose the parameters of neural network models and therefore the parameters are chosen according to what the researcher considers to be the best. Such an approach, however, implies that the risk of making bad decisions is extremely high which could explain why in many studies neural network models do not consistently perform better than their time series counterparts. In this paper through extensive experimentation the level of subjectivity in building neural network models is considerably reduced and therefore giving them a better chance of performing well. Our results show that in general neural network models perform better than traditionally used time series models in forecasting exchange rates.

Variational Markov Chain Monte Carlo for Bayesian smoothing of non-linear niffusions

In this paper we develop set of novel Markov chain Monte Carlo algorithms for Bayesian smoothing of partially observed non-linear diffusion processes. The sampling algorithms developed herein use a deterministic approximation to the posterior distribution over paths as the proposal distribution for a mixture of an independence and a random walk sampler. The approximating distribution is sampled by simulating an optimized time-dependent linear diffusion process derived from the recently developed variational Gaussian process approximation method. Flexible blocking strategies are introduced to further improve mixing, and thus the efficiency, of the sampling algorithms. The algorithms are tested on two diffusion processes: one with double-well potential drift and another with SINE drift. The new algorithm's accuracy and efficiency is compared with state-of-the-art hybrid Monte Carlo based path sampling. It is shown that in practical, finite sample, applications the algorithm is accurate except in the presence of large observation errors and low observation densities, which lead to a multi-modal structure in the posterior distribution over paths. More importantly, the variational approximation assisted sampling algorithm outperforms hybrid Monte Carlo in terms of computational efficiency, except when the diffusion process is densely observed with small errors in which case both algorithms are equally efficient.

Sequential, Bayesian geostatistics:a principled method for large data sets

The principled statistical application of Gaussian random field models used in geostatistics has historically been limited to data sets of a small size. This limitation is imposed by the requirement to store and invert the covariance matrix of all the samples to obtain a predictive distribution at unsampled locations, or to use likelihood-based covariance estimation. Various ad hoc approaches to solve this problem have been adopted, such as selecting a neighborhood region and/or a small number of observations to use in the kriging process, but these have no sound theoretical basis and it is unclear what information is being lost. In this article, we present a Bayesian method for estimating the posterior mean and covariance structures of a Gaussian random field using a sequential estimation algorithm. By imposing sparsity in a well-defined framework, the algorithm retains a subset of “basis vectors” that best represent the “true” posterior Gaussian random field model in the relative entropy sense. This allows a principled treatment of Gaussian random field models on very large data sets. The method is particularly appropriate when the Gaussian random field model is regarded as a latent variable model, which may be nonlinearly related to the observations. We show the application of the sequential, sparse Bayesian estimation in Gaussian random field models and discuss its merits and drawbacks.

Efficacy of memantine on behavioral and psychological symptoms related to dementia:a systematic meta-analysis

BACKGROUND: The behavioral and psychological symptoms related to dementia (BPSD) are difficult to manage and are associated with adverse patient outcomes. OBJECTIVE: To systematically analyze the data on memantine in the treatment of BPSD. METHODS: We searched MEDLINE, EMBASE, Pharm-line, the Cochrane Centre Collaboration, www.clinicaltrials.gov, www.controlled-trials.com, and PsycINFO (1966-July 2007). We contacted manufacturers and scrutinized the reference sections of articles identified in our search for further references, including conference proceedings. Two researchers (IM and CF) independently reviewed all studies identified by the search strategy. We included 6 randomized, parallel-group, double-blind studies that rated BPSD with the Neuropsychiatric Inventory (NPI) in our meta-analysis. Patients had probable Alzheimer's disease and received treatment with memantine for at least one month. Overall efficacy of memantine on the NPI was established with a t-test for the average difference between means across studies, using a random effects model. RESULTS: Five of the 6 studies identified had NPI outcome data. In these 5 studies, 868 patients were treated with memantine and 882 patients were treated with placebo. Patients on memantine improved by 1.99 on the NPI scale (95% Cl -0.08 to -3.91; p = 0.041) compared with the placebo group. CONCLUSIONS: Initial data appear to indicate that memantine decreases NPI scores and may have a role in managing BPSD. However, there are a number of limitations with the current data; the effect size was relatively small, and whether memantine produces significant clinical benefit is not clear.

A comparison of sales response predictions from demand models applied to store-level versus panel data

In order to generate sales promotion response predictions, marketing analysts estimate demand models using either disaggregated (consumer-level) or aggregated (store-level) scanner data. Comparison of predictions from these demand models is complicated by the fact that models may accommodate different forms of consumer heterogeneity depending on the level of data aggregation. This study shows via simulation that demand models with various heterogeneity specifications do not produce more accurate sales response predictions than a homogeneous demand model applied to store-level data, with one major exception: a random coefficients model designed to capture within-store heterogeneity using store-level data produced significantly more accurate sales response predictions (as well as better fit) compared to other model specifications. An empirical application to the paper towel product category adds additional insights. This article has supplementary material online.

Success factors of product innovation:an updated meta-analysis

Assessing factors that predict new product success (NPS) holds critical importance for companies, as research shows that despite considerable new product investment, success rates are generally below 25%. Over the decades, meta-analytical attempts have been made to summarize empirical findings on NPS factors. However, market environment changes such as increased global competition, as well as methodological advancements in meta-analytical research, present a timely opportunity to augment their results. Hence, a key objective of this research is to provide an updated and extended meta-analytic investigation of the factors affecting NPS. Using Henard and Szymanski's meta-analysis as the most comprehensive recent summary of empirical findings, this study updates their findings by analyzing articles published from 1999 through 2011, the period following the original meta-analysis. Based on 233 empirical studies (from 204 manuscripts) on NPS, with a total 2618 effect sizes, this study also takes advantage of more recent methodological developments by re-calculating effects of the meta-analysis employing a random effects model. The study's scope broadens by including overlooked but important additional variables, notably “country culture,” and discusses substantive differences between the updated meta-analysis and its predecessor. Results reveal generally weaker effect sizes than those reported by Henard and Szymanski in 2001, and provide evolutionary evidence of decreased effects of common success factors over time. Moreover, culture emerges as an important moderating factor, weakening effect sizes for individualistic countries and strengthening effects for risk-averse countries, highlighting the importance of further investigating culture's role in product innovation studies, and of tracking changes of success factors of product innovations. Finally, a sharp increase since 1999 in studies investigating product and process characteristics identifies a significant shift in research interest in new product development success factors. The finding that the importance of success factors generally declines over time calls for new theoretical approaches to better capture the nature of new product development (NPD) success factors. One might speculate that the potential to create competitive advantages through an understanding of NPD success factors is reduced as knowledge of these factors becomes more widespread among managers. Results also imply that managers attempting to improve success rates of NPDs need to consider national culture as this factor exhibits a strong moderating effect: Working in varied cultural contexts will result in differing antecedents of successful new product ventures.

Thermal processing of miscanthus, sugarcane bagasse, sugarcane trash and their acid hydrolysis residues

The research presented in this thesis was developed as part of DIBANET, an EC funded project aiming to develop an energetically self-sustainable process for the production of diesel miscible biofuels (i.e. ethyl levulinate) via acid hydrolysis of selected biomass feedstocks. Three thermal conversion technologies, pyrolysis, gasification and combustion, were evaluated in the present work with the aim of recovering the energy stored in the acid hydrolysis solid residue (AHR). Mainly consisting of lignin and humins, the AHR can contain up to 80% of the energy in the original feedstock. Pyrolysis of AHR proved unsatisfactory, so attention focussed on gasification and combustion with the aim of producing heat and/or power to supply the energy demanded by the ethyl levulinate production process. A thermal processing rig consisting on a Laminar Entrained Flow Reactor (LEFR) equipped with solid and liquid collection and online gas analysis systems was designed and built to explore pyrolysis, gasification and air-blown combustion of AHR. Maximum liquid yield for pyrolysis of AHR was 30wt% with volatile conversion of 80%. Gas yield for AHR gasification was 78wt%, with 8wt% tar yields and conversion of volatiles close to 100%. 90wt% of the AHR was transformed into gas by combustion, with volatile conversions above 90%. 5volO2%-95vol%N2 gasification resulted in a nitrogen diluted, low heating value gas (2MJ/m3). Steam and oxygen-blown gasification of AHR were additionally investigated in a batch gasifier at KTH in Sweden. Steam promoted the formation of hydrogen (25vol%) and methane (14vol%) improving the gas heating value to 10MJ/m3, below the typical for steam gasification due to equipment limitations. Arrhenius kinetic parameters were calculated using data collected with the LEFR to provide reaction rate information for process design and optimisation. Activation energy (EA) and pre-exponential factor (ko in s-1) for pyrolysis (EA=80kJ/mol, lnko=14), gasification (EA=69kJ/mol, lnko=13) and combustion (EA=42kJ/mol, lnko=8) were calculated after linearly fitting the data using the random pore model. Kinetic parameters for pyrolysis and combustion were also determined by dynamic thermogravimetric analysis (TGA), including studies of the original biomass feedstocks for comparison. Results obtained by differential and integral isoconversional methods for activation energy determination were compared. Activation energy calculated by the Vyazovkin method was 103-204kJ/mol for pyrolysis of untreated feedstocks and 185-387kJ/mol for AHRs. Combustion activation energy was 138-163kJ/mol for biomass and 119-158 for AHRs. The non-linear least squares method was used to determine reaction model and pre-exponential factor. Pyrolysis and combustion of biomass were best modelled by a combination of third order reaction and 3 dimensional diffusion models, while AHR decomposed following the third order reaction for pyrolysis and the 3 dimensional diffusion for combustion.

Unifractality and multifractality in the Italian stock market

Tests for random walk behaviour in the Italian stock market are presented, based on an investigation of the fractal properties of the log return series for the Mibtel index. The random walk hypothesis is evaluated against alternatives accommodating either unifractality or multifractality. Critical values for the test statistics are generated using Monte Carlo simulations of random Gaussian innovations. Evidence is reported of multifractality, and the departure from random walk behaviour is statistically significant on standard criteria. The observed pattern is attributed primarily to fat tails in the return probability distribution, associated with volatility clustering in returns measured over various time scales. © 2009 Elsevier Inc. All rights reserved.

Unfolding kernel embeddings of graphs:enhancing class separation through manifold learning

In this paper, we investigate the use of manifold learning techniques to enhance the separation properties of standard graph kernels. The idea stems from the observation that when we perform multidimensional scaling on the distance matrices extracted from the kernels, the resulting data tends to be clustered along a curve that wraps around the embedding space, a behavior that suggests that long range distances are not estimated accurately, resulting in an increased curvature of the embedding space. Hence, we propose to use a number of manifold learning techniques to compute a low-dimensional embedding of the graphs in an attempt to unfold the embedding manifold, and increase the class separation. We perform an extensive experimental evaluation on a number of standard graph datasets using the shortest-path (Borgwardt and Kriegel, 2005), graphlet (Shervashidze et al., 2009), random walk (Kashima et al., 2003) and Weisfeiler-Lehman (Shervashidze et al., 2011) kernels. We observe the most significant improvement in the case of the graphlet kernel, which fits with the observation that neglecting the locational information of the substructures leads to a stronger curvature of the embedding manifold. On the other hand, the Weisfeiler-Lehman kernel partially mitigates the locality problem by using the node labels information, and thus does not clearly benefit from the manifold learning. Interestingly, our experiments also show that the unfolding of the space seems to reduce the performance gap between the examined kernels.

Variational Markov chain Monte Carlo for Bayesian smoothing of non-linear diffusions

In this paper we develop set of novel Markov Chain Monte Carlo algorithms for Bayesian smoothing of partially observed non-linear diffusion processes. The sampling algorithms developed herein use a deterministic approximation to the posterior distribution over paths as the proposal distribution for a mixture of an independence and a random walk sampler. The approximating distribution is sampled by simulating an optimized time-dependent linear diffusion process derived from the recently developed variational Gaussian process approximation method. The novel diffusion bridge proposal derived from the variational approximation allows the use of a flexible blocking strategy that further improves mixing, and thus the efficiency, of the sampling algorithms. The algorithms are tested on two diffusion processes: one with double-well potential drift and another with SINE drift. The new algorithm's accuracy and efficiency is compared with state-of-the-art hybrid Monte Carlo based path sampling. It is shown that in practical, finite sample applications the algorithm is accurate except in the presence of large observation errors and low to a multi-modal structure in the posterior distribution over paths. More importantly, the variational approximation assisted sampling algorithm outperforms hybrid Monte Carlo in terms of computational efficiency, except when the diffusion process is densely observed with small errors in which case both algorithms are equally efficient. © 2011 Springer-Verlag.

Solvable model for distribution networks on random graphs

We propose a simple model that captures the salient properties of distribution networks, and study the possible occurrence of blackouts, i.e., sudden failings of large portions of such networks. The model is defined on a random graph of finite connectivity. The nodes of the graph represent hubs of the network, while the edges of the graph represent the links of the distribution network. Both, the nodes and the edges carry dynamical two state variables representing the functioning or dysfunctional state of the node or link in question. We describe a dynamical process in which the breakdown of a link or node is triggered when the level of maintenance it receives falls below a given threshold. This form of dynamics can lead to situations of catastrophic breakdown, if levels of maintenance are themselves dependent on the functioning of the net, once maintenance levels locally fall below a critical threshold due to fluctuations. We formulate conditions under which such systems can be analyzed in terms of thermodynamic equilibrium techniques, and under these conditions derive a phase diagram characterizing the collective behavior of the system, given its model parameters. The phase diagram is confirmed qualitatively and quantitatively by simulations on explicit realizations of the graph, thus confirming the validity of our approach. © 2007 The American Physical Society.

Persistence in a random bond ising model of socio-econo dynamics

We study the persistence phenomenon in a socio-econo dynamics model using computer simulations at a nite temperature on hypercubic lattices in dimensions up to ve. The model includes a \social" local eld which contains the magnetization at time t. The nearest neighbour quenched interactions are drawn from a binary distribution which is a function of the bond concentration, p. The decay of the persistence probability in the model depends on both the spatial dimension and p. We nd no evidence of \blocking" in this model. We also discuss the implications of our results for possible applications in the social and economic elds. It is suggested that the absence, or otherwise, of blocking could be used as a criterion to decide on the validity of a given model in dierent scenarios.

NLSE-based model of a random distributed feedback fiber laser

In this work we propose a NLSE-based model of power and spectral properties of the random distributed feedback (DFB) fiber laser. The model is based on coupled set of non-linear Schrödinger equations for pump and Stokes waves with the distributed feedback due to Rayleigh scattering. The model considers random backscattering via its average strength, i.e. we assume that the feedback is incoherent. In addition, this allows us to speed up simulations sufficiently (up to several orders of magnitude). We found that the model of the incoherent feedback predicts the smooth and narrow (comparing with the gain spectral profile) generation spectrum in the random DFB fiber laser. The model allows one to optimize the random laser generation spectrum width varying the dispersion and nonlinearity values: we found, that the high dispersion and low nonlinearity results in narrower spectrum that could be interpreted as four-wave mixing between different spectral components in the quasi-mode-less spectrum of the random laser under study could play an important role in the spectrum formation. Note that the physical mechanism of the random DFB fiber laser formation and broadening is not identified yet. We investigate temporal and statistical properties of the random DFB fiber laser dynamics. Interestingly, we found that the intensity statistics is not Gaussian. The intensity auto-correlation function also reveals that correlations do exist. The possibility to optimize the system parameters to enhance the observed intrinsic spectral correlations to further potentially achieved pulsed (mode-locked) operation of the mode-less random distributed feedback fiber laser is discussed.

A generative model for separating illumination and reflectance from images

It is well known that even slight changes in nonuniform illumination lead to a large image variability and are crucial for many visual tasks. This paper presents a new ICA related probabilistic model where the number of sources exceeds the number of sensors to perform an image segmentation and illumination removal, simultaneously. We model illumination and reflectance in log space by a generalized autoregressive process and Hidden Gaussian Markov random field, respectively. The model ability to deal with segmentation of illuminated images is compared with a Canny edge detector and homomorphic filtering. We apply the model to two problems: synthetic image segmentation and sea surface pollution detection from intensity images.