Forecasting under model uncertainty: Non-homogeneous hidden Markov models with Pòlya-Gamma data augmentation

We consider finite state-space non-homogeneous hidden Markov models for forecasting univariate time series. Given a set of predictors, the time series are modeled via predictive regressions with state-dependent coefficients and time-varying transition probabilities that depend on the predictors via a logistic/multinomial function. In a hidden Markov setting, inference for logistic regression coefficients becomes complicated and in some cases impossible due to convergence issues. In this paper, we aim to address this problem utilizing the recently proposed Pólya-Gamma latent variable scheme. Also, we allow for model uncertainty regarding the predictors that affect the series both linearly — in the mean — and non-linearly — in the transition matrix. Predictor selection and inference on the model parameters are based on an automatic Markov chain Monte Carlo scheme with reversible jump steps. Hence the proposed methodology can be used as a black box for predicting time series. Using simulation experiments, we illustrate the performance of our algorithm in various setups, in terms of mixing properties, model selection and predictive ability. An empirical study on realized volatility data shows that our methodology gives improved forecasts compared to benchmark models.     

An ensemble of LSTM neural networks for high‐frequency stock market classification

We propose an ensemble of long–short‐term memory (LSTM) neural networks for intraday stock predictions, using a large variety of technical analysis indicators as network inputs. The proposed ensemble operates in an online way, weighting the individual models proportionally to their recent performance, which allows us to deal with possible nonstationarities in an innovative way. The performance of the models is measured by area under the curve of the receiver operating characteristic. We evaluate the predictive power of our model on several US large‐cap stocks and benchmark it against lasso and ridge logistic classifiers. The proposed model is found to perform better than the benchmark models or equally weighted ensembles.     

Backward‐in‐Time Selection of the Order of Dynamic Regression Prediction Model

We investigate the optimal structure of dynamic regression models used in multivariate time series prediction and propose a scheme to form the lagged variable structure called Backward‐in‐Time Selection (BTS), which takes into account feedback and multicollinearity, often present in multivariate time series. We compare BTS to other known methods, also in conjunction with regularization techniques used for the estimation of model parameters, namely principal components, partial least squares and ridge regression estimation. The predictive efficiency of the different models is assessed by means of Monte Carlo simulations for different settings of feedback and multicollinearity. The results show that BTS has consistently good prediction performance, while other popular methods have varying and often inferior performance. The prediction performance of BTS was also found the best when tested on human electroencephalograms of an epileptic seizure, and for the prediction of returns of indices of world financial markets.Copyright © 2013 John Wiley & Sons, Ltd.     

Objectivity in confirmation: Post hoc monsters and novel predictions

The aim of this paper is to put in place some cornerstones in the foundations for an objective theory of confirmation by considering lessons from the failures of predictivism. Discussion begins with a widely accepted challenge, to find out what is needed in addition to the right kind of inferential–semantical relations between hypothesis and evidence to have a complete account of confirmation, one that gives a definitive answer to the question whether hypotheses branded as “post hoc monsters” can be confirmed. The predictivist view is then presented as a way to meet this challenge. Particular attention is paid to Worrall’s version of predictivism, as it appears to be the most sophisticated of the lot. It is argued that, despite its faults, his view turns our heads in the right direction by attempting to remove contingent considerations from confirmational matters. The demand to remove such considerations becomes the first of four cornerstones. Each cornerstone is put in place with the aim to steer clear of the sort of failures that plague various kinds of predictivism. In the process, it becomes obvious that the original challenge is wrongheaded and in need of revision. The paper ends with just such a revision.     

A genome-wide scan identifies mutations in the gene encoding phosphodiesterase 11A4 (PDE11A) in individuals with adrenocortical hyperplasia

Phosphodiesterases (PDEs) regulate cyclic nucleotide levels. Increased cyclic AMP (cAMP) signaling has been associated with PRKAR1A or GNAS mutations and leads to adrenocortical tumors and Cushing syndrome. We investigated the genetic source of Cushing syndrome in individuals with adrenocortical hyperplasia that was not caused by known defects. We performed genome-wide SNP genotyping, including the adrenocortical tumor DNA. The region with the highest probability to harbor a susceptibility gene by loss of heterozygosity (LOH) and other analyses was 2q31-2q35. We identified mutations disrupting the expression of the PDE11A isoform-4 gene (PDE11A) in three kindreds. Tumor tissues showed 2q31-2q35 LOH, decreased protein expression and high cyclic nucleotide levels and cAMP-responsive element binding protein (CREB) phosphorylation. PDE11A codes for a dual-specificity PDE that is expressed in adrenal cortex and is partially inhibited by tadalafil and other PDE inhibitors; its germline inactivation is associated with adrenocortical hyperplasia, suggesting another means by which dysregulation of cAMP signaling causes endocrine tumors.     

Are common fragile sites merely structural domains or highly organized “functional” units susceptible to oncogenic stress?

Common fragile sites (CFSs) are regions of the genome with a predisposition to DNA double-strand breaks in response to intrinsic (oncogenic) or extrinsic replication stress. CFS breakage is a common feature in carcinogenesis from its earliest stages. Given that a number of oncogenes and tumor suppressors are located within CFSs, a question that emerges is whether fragility in these regions is only a structural “passive” incident or an event with a profound biological effect. Furthermore, there is sparse evidence that other elements, like non-coding RNAs, are positioned with them. By analyzing data from various libraries, like miRbase and ENCODE, we show a prevalence of various cancer-related genes, miRNAs, and regulatory binding sites, such as CTCF within CFSs. We propose that CFSs are not only susceptible structural domains, but highly organized “functional” entities that when targeted, severe repercussion for cell homeostasis occurs.     

Encoding of conditioned fear in central amygdala inhibitory circuits

The central amygdala (CEA), a nucleus predominantly composed of GABAergic inhibitory neurons, is essential for fear conditioning. How the acquisition and expression of conditioned fear are encoded within CEA inhibitory circuits is not understood. Using in vivo electrophysiological, optogenetic and pharmacological approaches in mice, we show that neuronal activity in the lateral subdivision of the central amygdala (CEl) is required for fear acquisition, whereas conditioned fear responses are driven by output neurons in the medial subdivision (CEm). Functional circuit analysis revealed that inhibitory CEA microcircuits are highly organized and that cell-type-specific plasticity of phasic and tonic activity in the CEl to CEm pathway may gate fear expression and regulate fear generalization. Our results define the functional architecture of CEA microcircuits and their role in the acquisition and regulation of conditioned fear behaviour.     

ATP-sensitive potassium channels mediate survival during infection in mammals and insects

Specific homeostatic mechanisms confer stability in innate immune responses, preventing injury or death from infection. Here we identify, from a screen of N-ethyl-N-nitrosourea-mutagenized mice, a mutation causing both profound susceptibility to infection by mouse cytomegalovirus and approximately 20,000-fold sensitization to lipopolysaccharide (LPS), poly(I.C) and immunostimulatory (CpG) DNA. The LPS hypersensitivity phenotype is not suppressed by mutations in Myd88, Trif, Tnf, Tnfrsf1a, Ifnb, Ifng or Stat1, genes contributing to LPS responses, and results from an abnormality extrinsic to hematopoietic cells. The phenotype is due to a null allele of Kcnj8, encoding Kir6.1, a protein that combines with SUR2 to form an ATP-sensitive potassium channel (K(ATP)) expressed in coronary artery smooth muscle and endothelial cells. In Drosophila melanogaster, suppression of dSUR by RNA interference similarly causes hypersensitivity to infection by flock house virus. Thus, K(ATP) evolved to serve a homeostatic function during infection, and in mammals it prevents coronary artery vasoconstriction induced by cytokines dependent on TLR and/or MDA5 immunoreceptors.