Hierarchical Inference Research Articles

Explaining the trends and variability in the United States tornado records using climate teleconnections and shifts in observational practices

The annual frequency of tornadoes during 1950–2018 across the major tornado-impacted states were examined and modeled using anthropogenic and large-scale climate covariates in a hierarchical Bayesian inference framework. Anthropogenic factors include increases in population density and better detection systems since the mid-1990s. Large-scale climate variables include El Niño Southern Oscillation (ENSO), Southern Oscillation Index (SOI), North Atlantic Oscillation (NAO), Pacific Decadal Oscillation (PDO), Arctic Oscillation (AO), and Atlantic Multi-decadal Oscillation (AMO). The model provides a robust way of estimating the response coefficients by considering pooling of information across groups of states that belong to Tornado Alley, Dixie Alley, and Other States, thereby reducing their uncertainty. The influence of the anthropogenic factors and the large-scale climate variables are modeled in a nested framework to unravel secular trend from cyclical variability. Population density explains the long-term trend in Dixie Alley. The step-increase induced due to the installation of the Doppler Radar systems explains the long-term trend in Tornado Alley. NAO and the interplay between NAO and ENSO explained the interannual to multi-decadal variability in Tornado Alley. PDO and AMO are also contributing to this multi-time scale variability. SOI and AO explain the cyclical variability in Dixie Alley. This improved understanding of the variability and trends in tornadoes should be of immense value to public planners, businesses, and insurance-based risk management agencies.

Interpreting Bottom-Up Decision-Making of CNNs via Hierarchical Inference.

With the great success of convolutional neural networks (CNNs), interpretation of their internal network mechanism has been increasingly critical, while the network decision-making logic is still an open issue. In the bottom-up hierarchical logic of neuroscience, the decision-making process can be deduced from a series of sub-decision-making processes from low to high levels. Inspired by this, we propose the Concept-harmonized HierArchical INference (CHAIN) interpretation scheme. In CHAIN, a network decision-making process from shallow to deep layers is interpreted by the hierarchical backward inference based on visual concepts from high to low semantic levels. Firstly, we learned a general hierarchical visual-concept representation in CNN layered feature space by concept harmonizing model on a large concept dataset. Secondly, for interpreting a specific network decision-making process, we conduct the concept-harmonized hierarchical inference backward from the highest to the lowest semantic level. Specifically, the network learning for a target concept at a deeper layer is disassembled into that for concepts at shallower layers. Finally, a specific network decision-making process is explained as a form of concept-harmonized hierarchical inference, which is intuitively comparable to the bottom-up hierarchical visual recognition way. Quantitative and qualitative experiments demonstrate the effectiveness of the proposed CHAIN at both instance and class levels.

Usability in the Product Design Process Based on the Intelligent Microprocessor

The sharing and reuse of product design experience knowledge is beneficial to shorten the product development cycle and improve product quality. The intelligent microprocessor has greatly changed the way of interaction between people and computers. This article mainly studies the usability analysis in the product design process based on the intelligent microprocessor. The experiment must first perform initial finite element calculations, and the purpose is to complete the data collection. The collection of evaluation indicators is carried out with the help of perceptual vocabulary classification, and the hierarchical inference method is used to build a tree‐like analysis chart. After the indicator collection is completed, the Likert scale method is used to screen the indicators, and the evaluators will rate the selected indicators according to their importance. The Petri net modeling method is used to model the product design process, which mainly includes task element definition, task module division, task planning, and task allocation. Finally, the usability test of the product is carried out. Experimental data shows that the probability of completing the design task within 30 days is 27.74%. The results show that the intelligent microprocessor shortens the time‐consuming process of product design and at the same time enhances product usability.

Multi-task learning for data-efficient spatiotemporal modeling of tool surface progression in ultrasonic metal welding

Spatiotemporal processes commonly exist in manufacturing. Modeling and monitoring such processes are crucial for ensuring high-quality production. For example, ultrasonic metal welding is an important industrial-scale joining technique with wide applications. The surfaces of ultrasonic welding tools evolve in both spatial and temporal domains, resulting in a spatiotemporal process. Close monitoring of tool surface progression is imperative since degraded tools often lead to low-quality joints. However, it is generally expensive and time-consuming to acquire fine-scale surface measurement data, which is not economically viable. This paper develops a multi-task learning method to enable data-efficient spatiotemporal modeling. A Gaussian process-based hierarchical Bayesian inference structure is constructed to transfer knowledge among multiple similar-but-not-identical measurement tasks. Meanwhile, a spatiotemporal kernel is developed based on squared sine exponential damping (SSED) function to characterize the periodic trend of anvil surfaces. The proposed method is able to improve interpolation accuracy using limited measurement data compared with state-of-the-art techniques. Data collected from lithium-ion battery production are employed to demonstrate the effectiveness of the proposed method. Additionally, the influence of training data size and hyperparameter selection on the modeling performance is systematically investigated.

Cholinergic and dopaminergic effects on prediction error and uncertainty responses during sensory associative learning

Navigating the physical world requires learning probabilistic associations between sensory events and their change in time (volatility). Bayesian accounts of this learning process rest on hierarchical prediction errors (PEs) that are weighted by estimates of uncertainty (or its inverse, precision). In a previous fMRI study we found that low-level precision-weighted PEs about visual outcomes (that update beliefs about associations) activated the putative dopaminergic midbrain; by contrast, precision-weighted PEs about cue-outcome associations (that update beliefs about volatility) activated the cholinergic basal forebrain. These findings suggested selective dopaminergic and cholinergic influences on precision-weighted PEs at different hierarchical levels.Here, we tested this hypothesis, repeating our fMRI study under pharmacological manipulations in healthy participants. Specifically, we performed two pharmacological fMRI studies with a between-subject double-blind placebo-controlled design: study 1 used antagonists of dopaminergic (amisulpride) and muscarinic (biperiden) receptors, study 2 used enhancing drugs of dopaminergic (levodopa) and cholinergic (galantamine) modulation.Pooled across all pharmacological conditions of study 1 and study 2, respectively, we found that low-level precision-weighted PEs activated the midbrain and high-level precision-weighted PEs the basal forebrain as in our previous study. However, we found pharmacological effects on brain activity associated with these computational quantities only when splitting the precision-weighted PEs into their PE and precision components: in a brainstem region putatively containing cholinergic (pedunculopontine and laterodorsal tegmental) nuclei, biperiden (compared to placebo) enhanced low-level PE responses and attenuated high-level PE activity, while amisulpride reduced high-level PE responses. Additionally, in the putative dopaminergic midbrain, galantamine compared to placebo enhanced low-level PE responses (in a body-weight dependent manner) and amisulpride enhanced high-level precision activity. Task behaviour was not affected by any of the drugs.These results do not support our hypothesis of a clear-cut dichotomy between different hierarchical inference levels and neurotransmitter systems, but suggest a more complex interaction between these neuromodulatory systems and hierarchical Bayesian quantities. However, our present results may have been affected by confounds inherent to pharmacological fMRI. We discuss these confounds and outline improved experimental tests for the future.

Experts' opinion-based Bayesian Inference for the Coverage Factor in BOP Reliability

The blowout preventer is an important system for safety assurance in well drilling operations. Therefore, the efficient management of its reliability is an essential issue for risk prevention. In the present paper one considers a dynamic Bayesian network for reliability modeling, approaching corrective maintenance and an imperfect coverage factor, combining with a hierarchical Bayesian inference. Traditionally, the coverage factor in this kind of analysis for well safety is based on belief values. Considering that, we will approach it through a Bayesian analysis for the experts’ opinion. Thus, the present paper presents a fully Bayesian approach for blowout preventer reliability analysis, combining experts’ opinions in a dynamic Bayesian net modeling. In the end, is possible to compare results with and without imperfect coverage factor, and with belief values. The obtained results for a hypothetical case show the importance of considering the estimates’ uncertainties in the context of reliability management of drilling operations.

Hierarchical computing for hierarchical models in ecology

Abstract Bayesian hierarchical models allow ecologists to account for uncertainty and make inference at multiple scales. However, hierarchical models are often computationally intensive to fit, especially with large datasets, and researchers face trade‐offs between capturing ecological complexity in statistical models and implementing these models. We present a recursive Bayesian computing (RB) method that can be used to fit Bayesian models efficiently in sequential MCMC stages to ease computation and streamline hierarchical inference. We also introduce transformation‐assisted RB (TARB) to create unsupervised MCMC algorithms and improve interpretability of parameters. We demonstrate TARB by fitting a hierarchical animal movement model to obtain inference about individual‐ and population‐level migratory characteristics. Our recursive procedure reduced computation time for fitting our hierarchical movement model by half compared to fitting the model with a single MCMC algorithm. We obtained the same inference fitting our model using TARB as we obtained fitting the model with a single algorithm. For complex ecological statistical models, like those for animal movement, multi‐species systems, or large spatial and temporal scales, the computational demands of fitting models with conventional computing techniques can limit model specification, thus hindering scientific discovery. Transformation‐assisted RB is one of the most accessible methods for reducing these limitations, enabling us to implement new statistical models and advance our understanding of complex ecological phenomena.

Simultaneous Customers and Supplier’s Prioritization: An AHP-Based Fuzzy Inference Decision Support System (AHP-FIDSS)

Traditionally in supply chain management suppliers and customers are evaluated independently. This is in spite of the fact that supplier performance has direct impact on the performance of the customer. This research develops a mechanism for simultaneous evaluation and prioritization of supplier and customer using the sustainability metrics comprising economic, environment and social measures. In order to quantify the qualitative factors, priority index is introduced to measure the prioritization of customers and suppliers simultaneously. To get the priority index for each customer and supplier, a novel analytical hierarchical process-based fuzzy inference decision support system (AHP-FIDSS) has been introduced. An AHP-FIDSS involves the factor screening, hierarchical structure modeling, quantification of qualitative factors, and their conversion to quantitative values. AHP-FIDSS is knowledge-based system involving expert’s decision alternatives or logical rules. The number of input variables and their levels are proportional to the number of logical rules and rule size. In order to reduce the numbers of rules, a Taguchi orthogonal array is used that reduces the numbers of rules, thereby substantially simplifying the evaluation process A case study of a supply chain of surgical instruments has been presented for the real-time application of the proposed model. The results exhibited the simultaneous classification of customers and suppliers with respect to their priority index. Higher value of priority index indicates higher importance and vice versa. This research is useful for the supply chain mangers in procurement, sales and production to develop an importance classification for customers and suppliers in a supply chain.

Deep Active Inference and Scene Construction.

Adaptive agents must act in intrinsically uncertain environments with complex latent structure. Here, we elaborate a model of visual foraging—in a hierarchical context—wherein agents infer a higher-order visual pattern (a “scene”) by sequentially sampling ambiguous cues. Inspired by previous models of scene construction—that cast perception and action as consequences of approximate Bayesian inference—we use active inference to simulate decisions of agents categorizing a scene in a hierarchically-structured setting. Under active inference, agents develop probabilistic beliefs about their environment, while actively sampling it to maximize the evidence for their internal generative model. This approximate evidence maximization (i.e., self-evidencing) comprises drives to both maximize rewards and resolve uncertainty about hidden states. This is realized via minimization of a free energy functional of posterior beliefs about both the world as well as the actions used to sample or perturb it, corresponding to perception and action, respectively. We show that active inference, in the context of hierarchical scene construction, gives rise to many empirical evidence accumulation phenomena, such as noise-sensitive reaction times and epistemic saccades. We explain these behaviors in terms of the principled drives that constitute the expected free energy, the key quantity for evaluating policies under active inference. In addition, we report novel behaviors exhibited by these active inference agents that furnish new predictions for research on evidence accumulation and perceptual decision-making. We discuss the implications of this hierarchical active inference scheme for tasks that require planned sequences of information-gathering actions to infer compositional latent structure (such as visual scene construction and sentence comprehension). This work sets the stage for future experiments to investigate active inference in relation to other formulations of evidence accumulation (e.g., drift-diffusion models) in tasks that require planning in uncertain environments with higher-order structure.

HiGCIN: Hierarchical Graph-Based Cross Inference Network for Group Activity Recognition.

Group activity recognition (GAR) is a challenging task aimed at recognizing the behavior of a group of people. It is a complex inference process in which visual cues collected from individuals are integrated into the final prediction, being aware of the interaction between them. This paper goes one step further beyond the existing approaches by designing a Hierarchical Graph-based Cross Inference Network (HiGCIN), in which three levels of information, i.e., the body-region level, person level, and group-activity level, are constructed, learned, and inferred in an end-to-end manner. Primarily, we present a generic Cross Inference Block (CIB), which is able to concurrently capture the latent spatiotemporal dependencies among body regions and persons. Based on the CIB, two modules are designed to extract and refine features for group activities at each level. Experiments on two popular benchmarks verify the effectiveness of our approach, particularly in the ability to infer with multilevel visual cues. In addition, training our approach does not require individual action labels to be provided, which greatly reduces the amount of labor required in data annotation.

Fully Dynamic Inference with Deep Neural Networks

Modern deep neural networks are powerful and widely applicable models that extract task-relevant information through multi-level abstraction. Their cross-domain success, however, is often achieved at the expense of computational cost, high memory bandwidth, and long inference latency, which prevents their deployment in resource-constrained and time-sensitive scenarios, such as edge-side inference and self-driving cars. While recently developed methods for creating efficient deep neural networks are making their real-world deployment more feasible by reducing model size, they do not fully exploit input properties on a per-instance basis to maximize computational efficiency and task accuracy. In particular, most existing methods typically use a one-size-fits-all approach that identically processes all inputs. Motivated by the fact that different images require different feature embeddings to be accurately classified, we propose a fully dynamic paradigm that imparts deep convolutional neural networks with hierarchical inference dynamics at the level of layers and individual convolutional filters/channels. Two compact networks, called Layer-Net (L-Net) and Channel-Net (C-Net), predict on a per-instance basis which layers or filters/channels are redundant and therefore should be skipped. L-Net and C-Net also learn how to scale retained computation outputs to maximize task accuracy. By integrating L-Net and C-Net into a joint design framework, called LC-Net, we consistently outperform state-of-the-art dynamic frameworks with respect to both efficiency and classification accuracy. On the CIFAR-10 dataset, LC-Net results in up to <inline-formula><tex-math notation="LaTeX">$11.9\times$</tex-math><alternatives><mml:math><mml:mrow><mml:mn>11</mml:mn><mml:mo>.</mml:mo><mml:mn>9</mml:mn><mml:mo>&#x00D7;</mml:mo></mml:mrow></mml:math><inline-graphic xlink:href="xia-ieq1-3056031.gif"/></alternatives></inline-formula> fewer floating-point operations (FLOPs) and up to 3.3 percent higher accuracy compared to other dynamic inference methods. On the ImageNet dataset, LC-Net achieves up to <inline-formula><tex-math notation="LaTeX">$1.4\times$</tex-math><alternatives><mml:math><mml:mrow><mml:mn>1</mml:mn><mml:mo>.</mml:mo><mml:mn>4</mml:mn><mml:mo>&#x00D7;</mml:mo></mml:mrow></mml:math><inline-graphic xlink:href="xia-ieq2-3056031.gif"/></alternatives></inline-formula> fewer FLOPs and up to 4.6 percent higher Top-1 accuracy than the other methods.

Bayesian calibration of hysteretic reduced order structural models for earthquake engineering applications

Seismic risk assessment entails frequently a large number of nonlinear time-history analyses. The use of high-fidelity Finite Element Models (FEMs) in this context, though supports high accuracy predictions, involves a substantial computational burden. This has incentivized researchers to explore the use of calibrated Reduced Order Models (ROMs) to replace the computational expensive FEMs, with the calibration process established so that the ROM can match the desired FEM characteristics. Recent studies have explicitly examined this calibration for hysteretic, multi-degree-of-freedom ROMs utilizing FEM time-history response data. This calibration has been posed as a least-squares optimization problem for the configurable ROM parameters, considering the discrepancy between the ROM and FEM responses for a selected number of excitations, for different output quantities of interest (QoIs). This work extends these efforts to examine a Bayesian calibration framework of ROMs for earthquake engineering applications. The entire posterior distribution of the ROM parameters is identified, instead of solely the point estimate that the current least-squares optimization provides. This facilitates a quantification of the uncertainty in the calibrated ROM parameters. For formulating the likelihood function different assumptions are discussed, impacting, ultimately, how the different earthquake excitations and QoIs are effectively combined. Comparisons to the objective function utilized in the established least-squares approach are drawn. Furthermore, a model class selection is advocated to compare different candidate ROMs in order to choose the most appropriate one given the available data. Finally, a hierarchical Bayesian inference approach is examined, treating independently the ROM calibration for each earthquake excitation, but unifying the ROM parameters under a common prior, updated using data from all excitations. Different challenges are discussed for implementing such a multilevel Bayesian framework for the current application. The illustrative implementation considers two different structures, corresponding to different materials and heights, with the FEMs developed in OpenSees. The different calibration approaches are evaluated by utilizing the calibrated ROMs for seismic risk assessment for different seismicity scenarios.

Predictive sensorimotor control in autism.

Autism spectrum disorder has been characterized by atypicalities in how predictions and sensory information are processed in the brain. To shed light on this relationship in the context of sensorimotor control, we assessed prediction-related measures of cognition, perception, gaze and motor functioning in a large general population (n = 92; Experiment 1) and in clinically diagnosed autistic participants (n = 29; Experiment 2). In both experiments perception and action were strongly driven by prior expectations of object weight, with large items typically predicted to weigh more than equally-weighted smaller ones. Interestingly, these predictive action models were used comparably at a sensorimotor level in both autistic and neurotypical individuals with varying levels of autistic-like traits. Specifically, initial fingertip force profiles and resulting action kinematics were both scaled according to participants' pre-lift heaviness estimates, and generic visual sampling behaviours were notably consistent across groups. These results suggest that the weighting of prior information is not chronically underweighted in autism, as proposed by simple Bayesian accounts of the disorder. Instead, our results cautiously implicate context-sensitive processing mechanisms, such as precision modulation and hierarchical volatility inference. Together, these findings present novel implications for both future scientific investigations and the autism community.

Measuring angular N -point correlations of binary black hole merger gravitational-wave events with hierarchical Bayesian inference

Advanced LIGO and Virgo have detected ten binary black hole mergers by the end of their second observing run. These mergers have already allowed constraints to be placed on the population distribution of black holes in the Universe, which will only improve with more detections and increasing sensitivity of the detectors. In this paper we develop techniques to measure the angular distribution of black hole mergers by measuring their statistical N-point correlations through hierarchical Bayesian inference. We apply it to the special case of two-point angular correlations using a Legendre polynomial basis on the sky. Building on the mixture model formalism introduced in Ref.[1] we show how one can measure two-point correlations with no threshold on significance, allowing us to target the ensemble of sub-threshold binary black hole mergers not resolvable with the current generation of ground based detectors. We also show how one can use these methods to correlate gravitational waves with other probes of large scale angular structure like galaxy counts, and validate both techniques through simulations.

A Flexible and Explainable Vehicle Motion Prediction and Inference Framework Combining Semi-Supervised AOG and ST-LSTM

Accurate trajectory prediction of surrounding vehicles is important for automated vehicles. To solve several existing problems of maneuver-based trajectory prediction, we propose four targeted solutions and establish a trajectory prediction model that integrates semi-supervised And-or Graph (AOG) and Spatio-temporal LSTM (ST-LSTM). To reduce the dependence on the well-labeled dataset, we introduce the concept of sub-maneuvers to improve the classifications of vehicle movements based on the given rough maneuver labels. AOG is used as the backbone of the probabilistic motion inference considering sub-maneuvers. We only define the basic units and inference logics of AOG and design a semi-supervised approach to directly learn the sub-maneuvers and the inference model structure from the training data, without manually specifying the structure (layers and nodes) of the inference model. This approach helps to avoid excessive artificial design or biases. The learned hierarchical motion inference model improves the interpretability of the overall trajectory prediction process. To utilize vehicle interaction information and further yield more accurate prediction, we adopt two different methods to consider vehicle interaction in the two sub-models (maneuver recognition and trajectory prediction). The experiment on NGSIM I-80 dataset shows that the maneuver-based model proposed in this paper (AOG-ST and refined AOG-ST-TB) performs more accurate trajectory prediction results. Although the AOG-ST seems clumsy and slow, we show that it is a flexible and quick model for trajectory prediction for various driving scenarios through the discussion and experiment.

Detection and parameter estimation of binary neutron star merger remnants

Detection and parameter estimation of binary neutron star merger remnants can shed light on the physics of hot matter at supranuclear densities. Here we develop a fast, simple model that can generate gravitational waveforms, and show it can be used for both detection and parameter estimation of post-merger remnants. The model consists of three exponentially-damped sinusoids with a linear frequency-drift term. The median fitting factors between the model waveforms and numerical-relativity simulations exceed 0.90. We detect remnants at a post-merger signal-to-noise ratio of $\ge 7$ using a Bayes-factor detection statistic with a threshold of 3000. We can constrain the primary post-merger frequency to $\pm_{1.2}^{1.4}\%$ at post-merger signal-to-noise ratios of 15 with an increase in precision to $\pm_{0.2}^{0.3}\%$ for post-merger signal-to-noise ratios of 50. The tidal coupling constant can be constrained to $\pm^{9}_{12}\%$ at post-merger signal-to-noise ratios of 15, and $\pm 5\%$ at post-merger signal-to-noise ratios of 50 using a hierarchical inference model.

The cosmic evolution of the stellar mass–velocity dispersion relation of early-type galaxies

ABSTRACT We study the evolution of the observed correlation between central stellar velocity dispersion σe and stellar mass M* of massive ($M_*\gtrsim 3\times 10^{10}\, \mathrm{M_\odot}$) early-type galaxies (ETGs) out to redshift z ≈ 2.5, taking advantage of a Bayesian hierarchical inference formalism. Collecting ETGs from state-of-the-art literature samples, we build a fiducial sample (0 ≲ z ≲ 1), which is obtained with homogeneous selection criteria, but also a less homogeneous extended sample (0 ≲ z ≲ 2.5). Based on the fiducial sample, we find that at z ≲ 1 the M*–σe relation is well represented by $\sigma _{\mathrm{e}}\propto M_*^{\beta }(1+z)^{\zeta}$, with β ≃ 0.18 independent of redshift and ζ ≃ 0.4 (at a given M*, σe decreases for decreasing z, for instance by a factor of ≈1.3 from z = 1 to z = 0). When the slope β is allowed to evolve, we find it increasing with redshift: β(z) ≃ 0.16 + 0.26log (1 + z) describes the data as well as constant β ≃ 0.18. The intrinsic scatter of the M*–σe relation is ≃0.08 dex in σe at given M*, independent of redshift. Our results suggest that, on average, the velocity dispersion of individual massive (M* ≳ 3 × 1011M⊙) ETGs decreases with time while they evolve from z ≈ 1 to z ≈ 0. The analysis of the extended sample, over the wider redshift range 0 ≲ z ≲ 2.5, leads to results similar to that of the fiducial sample, with slightly stronger redshift dependence of the normalization (ζ ≃ 0.5) and weaker redshift dependence of the slope (dβ/dlog (1 + z) ≃ 0.18) when β varies with time. At z = 2 ETGs with $M_*\approx 10^{11}\, \mathrm{M_\odot}$ have, on average, ≈1.7 higher σe than ETGs of similar stellar mass at z = 0.

Mapping aboveground biomass and its prediction uncertainty using LiDAR and field data, accounting for tree-level allometric and LiDAR model errors

BackgroundThe increasing availability of remotely sensed data has recently challenged the traditional way of performing forest inventories, and induced an interest in model-based inference. Like traditional design-based inference, model-based inference allows for regional estimates of totals and means, but in addition for wall-to-wall mapping of forest characteristics. Recently Light Detection and Ranging (LiDAR)-based maps of forest attributes have been developed in many countries and been well received by users due to their accurate spatial representation of forest resources. However, the correspondence between such mapping and model-based inference is seldom appreciated. In this study we applied hierarchical model-based inference to produce aboveground biomass maps as well as maps of the corresponding prediction uncertainties with the same spatial resolution. Further, an estimator of mean biomass at regional level, and its uncertainty, was developed to demonstrate how mapping and regional level assessment can be combined within the framework of model-based inference.ResultsThrough a new version of hierarchical model-based estimation, allowing models to be nonlinear, we accounted for uncertainties in both the individual tree-level biomass models and the models linking plot level biomass predictions with LiDAR metrics. In a 5005 km2 large study area in south-central Sweden the predicted aboveground biomass at the level of 18 m ×18 m map units was found to range between 9 and 447 Mg ·ha−1. The corresponding root mean square errors ranged between 10 and 162 Mg ·ha−1. For the entire study region, the mean aboveground biomass was 55 Mg ·ha−1 and the corresponding relative root mean square error 8%. At this level 75% of the mean square error was due to the uncertainty associated with tree-level models.ConclusionsThrough the proposed method it is possible to link mapping and estimation within the framework of model-based inference. Uncertainties in both tree-level biomass models and models linking plot level biomass with LiDAR data are accounted for, both for the uncertainty maps and the overall estimates. The development of hierarchical model-based inference to handle nonlinear models was an important prerequisite for the study.

Benefits of commitment in hierarchical inference.

Humans have the tendency to commit to a single interpretation of what has caused some observed evidence rather than considering all possible alternatives. This tendency can explain various forms of biases in cognition and perception. However, committing to a single high-level interpretation seems short-sighted and irrational, and thus it is unclear why humans are motivated to use such strategy. In a first step toward answering this question, we systematically quantified how this strategy affects estimation accuracy at the feature level in the context of 2 common hierarchical inference tasks, category-based perception and causal cue combination. Using model simulations, we demonstrate that although estimation accuracy is generally impaired when conditioned on only a single high-level interpretation, the reduction is not uniform across the entire feature range. Compared with a full inference strategy that considers all high-level interpretations, accuracy is only worse for feature values relatively close to the decision boundaries but is better everywhere else. That is, for feature values for which an observer has a reasonably high chance of being correct about the high-level interpretation of the feature, a full commitment to that particular interpretation is advantageous. We also show that conditioning on an preceding high-level interpretation provides an effective mechanism for partially protecting the evidence from corruption with late noise in the inference process (e.g., during retention in and recall from working memory). Our results suggest that a top-down inference strategy that solely relies on the most likely high-level interpretation can be favorable with regard to late noise and more holistic performance metrics. (PsycInfo Database Record (c) 2020 APA, all rights reserved).

Precision weighting of cortical unsigned prediction error signals benefits learning, is mediated by dopamine, and is impaired in psychosis

Recent theories of cortical function construe the brain as performing hierarchical Bayesian inference. According to these theories, the precision of prediction errors plays a key role in learning and decision-making, is controlled by dopamine and contributes to the pathogenesis of psychosis. To test these hypotheses, we studied learning with variable outcome-precision in healthy individuals after dopaminergic modulation with a placebo, a dopamine receptor agonist bromocriptine or a dopamine receptor antagonist sulpiride (dopamine study n = 59) and in patients with early psychosis (psychosis study n = 74: 20 participants with first-episode psychosis, 30 healthy controls and 24 participants with at-risk mental state attenuated psychotic symptoms). Behavioural computational modelling indicated that precision weighting of prediction errors benefits learning in health and is impaired in psychosis. FMRI revealed coding of unsigned prediction errors, which signal surprise, relative to their precision in superior frontal cortex (replicated across studies, combined n = 133), which was perturbed by dopaminergic modulation, impaired in psychosis and associated with task performance and schizotypy (schizotypy correlation in 86 healthy volunteers). In contrast to our previous work, we did not observe significant precision-weighting of signed prediction errors, which signal valence, in the midbrain and ventral striatum in the healthy controls (or patients) in the psychosis study. We conclude that healthy people, but not patients with first-episode psychosis, take into account the precision of the environment when updating beliefs. Precision weighting of cortical prediction error signals is a key mechanism through which dopamine modulates inference and contributes to the pathogenesis of psychosis.