Towards end-to-end likelihood-free inference with convolutional neural networks

Complex simulator-based models with non-standard sampling distributions require sophisticated design choices for reliable approximate parameter inference. We introduce a fast, end-to-end approach for approximate Bayesian computation (ABC) based on fully convolutional neural networks. The method enables users of ABC to derive simultaneously the posterior mean and variance of multidimensional posterior distributions directly from raw simulated data. Once trained on simulated data, the convolutional neural network is able to map real data samples of variable size to the first two posterior moments of the relevant parameter's distributions. Thus, in contrast to other machine learning approaches to ABC, our approach allows us to generate reusable models that can be applied by different researchers employing the same model. We verify the utility of our method on two common statistical models (i.e., a multivariate normal distribution and a multiple regression scenario), for which the posterior parameter distributions can be derived analytically. We then apply our method to recover the parameters of the leaky competing accumulator (LCA) model and we reference our results to the current state-of-the-art technique, which is the probability density estimation (PDA). Results show that our method exhibits a lower approximation error compared with other machine learning approaches to ABC. It also performs similarly to PDA in recovering the parameters of the LCA model.     

A new quantile estimator with weights based on a subsampling approach

Quantiles are widely used in both theoretical and applied statistics, and it is important to be able to deploy appropriate quantile estimators. To improve performance in the lower and upper quantiles, especially with small sample sizes, a new quantile estimator is introduced which is a weighted average of all order statistics. The new estimator, denoted NO, has desirable asymptotic properties. Moreover, it offers practical advantages over four estimators in terms of efficiency in most experimental settings. The Harrell–Davis quantile estimator, the default quantile estimator of the R programming language, the Sfakianakis–Verginis SV2 quantile estimator and a kernel quantile estimator. The NO quantile estimator is also utilized in comparing two independent groups with a percentile bootstrap method and, as expected, it is more successful than other estimators in controlling Type I error rates.     

An Examination of the Biblical Evidence for Open Theism

Open Theist theologians argue that their view of divine foreknowledge and providence is the correct interpretation of the Bible, and suggest some biblical evidence to support this claim. Among these theologians, Gregory A. Boyd’s case is the most systematic, and also the most comprehensive and rigorous. Taking into consideration (1) the main philosophical claims of Open Theism and its main rivals, namely Calvinism and Molinism, and (2) Open Theist theologians’ interpretative principles for the Bible, the biblical evidence for Open Theism suggested by Boyd can be classified into three groups according to their strength. I will argue that the first group of themes has no evidential value for Open Theism, since these themes can be interpreted just as plausibly from a Calvinist or Molinist perspective. By contrast, the second group of themes has some evidential value and thus constitutes prima facie evidence for Open Theism, since these themes make most sense under an Open Theist interpretation. However, these themes also make some sense on a Molinist reading. The third group of themes, I argue, has most evidential value for Open Theism, since it seems hard to reconcile these themes either with Calvinism or Molinism.     

Betterness of permissibility

Philosophical Studies - It is often assumed that morally permissible acts are morally better than impermissible acts. We call this claim Betterness of Permissibility. Yet, we show that some...     

Effects of cognitive behavioural therapy and return-to-work intervention for patients on sick leave due to stress-related disorders: Results from a randomized trial

The aim of this study was to evaluate specific effects for patients with adjustment or exhaustion disorder, the Stress subgroup (n = 152), regarding symptom severity and sick leave after CBT, a return-to-work intervention (RTW-I), and a combination of them (COMBO), using data from a randomized trial. In the original study, primary care patients on sick leave (N = 211) were randomized to CBT (n = 64), RTW-I (n = 67), or COMBO (n = 80). Blinded Clinician Severity Rating (CSR) of symptoms and sick leave registry data were primary outcomes. Subgroup analyses showed that for the Stress subgroup, CBT led to greater reduction of symptoms than RTW-I posttreatment, but COMBO did not differ from CBT or RTW-I. Regarding sick leave, there was no difference between treatments in the Stress subgroup. An exploratory analysis of the treatment effects in a subgroup of patients with depression, anxiety or insomnia indicates that RTW-I reduced sick leave faster than CBT. We conclude that CBT may be promising as an effective treatment of stress and exhaustion disorder.     

Is backing grounding?

Separatists are grounding theorists who hold that grounding relations and metaphysical explanations are distinct, yet intimately connected in the sense that grounding relations back metaphysical explanations, just as causal relations back causal explanations. But Separatists have not elaborated on the nature of the ‘backing’ relation. In this paper, I argue that backing is a form of (partial) grounding. In particular, backing has many of the properties commonly attributed to grounding, and taking backing to be partial grounding allows Separatists to make the most of their position vis-à-vis their Unionist opponents.     

Instrumental Robots

Science and Engineering Ethics - Advances in artificial intelligence research allow us to build fairly sophisticated agents: robots and computer programs capable of acting and deciding on their own...     

Task concentration training versus applied relaxation, in combination with cognitive therapy, for social phobia patients with fear of blushing, trembling, and sweating

Social phobia patients with fear of blushing, trembling, sweating and/or freezing as main complaint (N = 65) were randomly assigned to either task concentration training (TCT) or applied relaxation (AR) both followed by cognitive therapy (CT). Measurements took place before and after wait-list, after TCT or AR (within-test), after CT (post-test), at 3-months and at 1-year follow-up. Effects were assessed on fear of showing bodily symptoms (the central outcome variable), social phobia, other psychopathology, social skills, self-consciousness, self-focused attention, and dysfunctional beliefs. No changes occurred during wait-list. Both treatments were highly effective. TCT was superior to AR in reducing fear of bodily symptoms and dysfunctional beliefs at within-test. This difference disappeared after CT, at post-test and at 3-months follow-up. However, at 1-year follow-up the combination TCT-CT was superior to AR-CT in reducing fear of bodily symptoms, and effect sizes for TCT-CT reached 3. Furthermore, at all assessment moments TCT or the combination TCT-CT was superior to AR-CT in reducing self-consciousness and self-focused attention. The superior long-term effect of TCT on fear of showing bodily symptoms is explained by lasting changes in attentional focus.     

Spatial frequency processing in schizophrenia: trait or state marker?

B. F. O'Donnell et al. found impaired discrimination performances at low and medium spatial frequencies in patients with schizophrenia. In this study, the authors replicated this finding in a group of remitted, unmedicated, and highly functioning outpatients with spared IQ and attentional functions. However, the deficit was restricted to low spatial frequencies (0.5 cycles/degree), which suggests that this deficit is a trait marker of schizophrenia.