Forward Model Predictions Research Articles

Facilitation of sensorimotor temporal recalibration mechanisms by cerebellar tDCS in patients with schizophrenia spectrum disorders and healthy individuals

Core symptoms in patients with schizophrenia spectrum disorders (SSD), like hallucinations or ego-disturbances, have been associated with a failure of internal forward models to predict the sensory outcomes of self-generated actions. Importantly, forward model predictions must also be able to flexibly recalibrate to changing environmental conditions, for example to account for additional delays between action and outcome. We investigated whether transcranial direct current stimulation (tDCS) can be used to improve these sensorimotor temporal recalibration mechanisms in patients and healthy individuals. While receiving tDCS on the cerebellum, temporo-parietal junction, supplementary motor area, or sham stimulation, patients with SSD and healthy control participants were repeatedly exposed to delays between actively or passively elicited button presses and auditory outcomes. Effects of this procedure on temporal perception were assessed with a delay detection task. Similar recalibration outcomes and faciliatory effects of cerebellar tDCS on recalibration were observed in SSD and healthy individuals. Our findings indicate that sensorimotor recalibration mechanisms may be preserved in SSD and highlight the importance of the cerebellum in both patients and healthy individuals for this process. They further suggest that cerebellar tDCS could be a promising tool for addressing deficits in action-outcome monitoring and related adaptive sensorimotor processes in SSD.

Dynamic Assimilation of Deep Learning Predictions to a Process-Based Water Budget

A three-step data assimilation (DA) of deep learning (DL) predictions to a process-based water budget is developed and applied to produce an active, operational water balance for groundwater management. In the first step, an existing water budget model provides forward model predictions of aquifer storage from meteorological observations, estimates of pumping and diversion discharge, and estimates of recharge. A Kalman filter DA approach is the second step and generates updated storage volumes by combining a long short-term memory (LSTM) network, a DL method, and predicted “measurements” with forward model predictions. The third “correction” step uses modified recharge and pumping, adjusted to account for the difference between Kalman update storage and forward model predicted storage, in forward model re-simulation to approximate updated storage volume. Use of modified inputs in the correction provides a mass-conservative water budget framework that leverages DL predictions. LSTM predictor “measurements” primarily represent missing observations due to missing or malfunctioning equipment. Pumping and recharge inputs are uncertain and unobserved in the study region and can be adjusted without contradicting measurements. Because DL requires clean and certain data for learning, a common-sense baseline facilitates interpretation of LSTM generalization skill and accounts for feature and outcome uncertainty when sufficient target data are available. DA, in contrast to DL, provides for explicit uncertainty analysis through an observation error model, which allows the integrated approach to address uncertainty impacts from an LSTM predictor developed from limited outcome observations.

Perception of self-generated and externally-generated visual stimuli: Evidence from EEG and behavior.

Efference copy-based forward model mechanisms may help us to distinguish between self-generated and externally-generated sensory consequences. Previous studies have shown that self-initiation modulates neural and perceptual responses to identical stimulation. For example, event-related potentials (ERPs) elicited by tones that follow a button press are reduced in amplitude relative to ERPs elicited by passively attended tones. However, previous EEG studies investigating visual stimuli in this context are rare, provide inconclusive results, and lack adequate control conditions with passive movements. Furthermore, although self-initiation is known to modulate behavioral responses, it is not known whether differences in the amplitude of ERPs also reflect differences in perception of sensory outcomes. In this study, we presented to participants visual stimuli consisting of gray discs following either active button presses, or passive button presses, in which an electromagnet moved the participant's finger. Two discs presented visually 500-1250 ms apart followed each button press, and participants judged which of the two was more intense. Early components of the primary visual response (N1 and P2) over the occipital electrodes were suppressed in the active condition. Interestingly, suppression in the intensity judgment task was only correlated with suppression of the visual P2 component. These data support the notion of efference copy-based forward model predictions in the visual sensory modality, but especially later processes (P2) seem to be perceptually relevant. Taken together, the results challenge the assumption that N1 differences reflect perceptual suppression and emphasize the relevance of the P2 ERP component.

The implications of an aging population and increased obesity for knee arthroplasty rates in Sweden: a register-based study

Background and purpose — Total knee arthroplasty (TKA) has increased substantially in Sweden. We quantified the relative risk for TKA in the Swedish population for different BMI categories and age groups to investigate whether the continued increase in TKA is attributable to increased prevalence of obesity and elderly people in the population, and to put forward model predictions for coming needs for TKA. Patients and methods — We used the Swedish Nationwide Health Survey (SNHS) and the Swedish Knee Arthroplasty Register (SKAR) 2009–2015 to calculate the relative risk (RR) of TKA by age (middle-aged 45–64 years and elderly 65–84 years) and BMI (BMI 18.5–24.9 normal weight; BMI 25.0–29.9 overweight; BMI > 30 obese). The RR for TKA was applied to the demographic forecasts for the Swedish population as a forecasting model. Results — Population size increased 5.2% from 2009 to 2015 to 40,000 middle-aged and 250,000 elderly, and the prevalence of obesity increased from 16% to 18% in these 2 age categories. Compared with those of normal weight, the RR for TKA was 2.7 (95% CI 2.5–3.0) higher for the overweight and 7.3 (6.7–8.0) higher for the obese, aged 45–64. The corresponding figures for individuals aged 65–84 were 2.1 (2.0–2.2) and 4.0 (3.8–4.3) higher, respectively. The changes in the prevalence of obesity and an increase in the elderly population accounted for an estimated increase of 1,700 TKAs over the 7 years. Interpretation — The increase in obesity frequency and the rise in the population of middle-aged and elderly may, to some extent, explain the rise in TKA utilization in Sweden.

A Cerebellum-Inspired Learning Approach for Adaptive and Anticipatory Control.

The cerebellum, which is responsible for motor control and learning, has been suggested to act as a Smith predictor for compensation of time-delays by means of internal forward models. However, insights about how forward model predictions are integrated in the Smith predictor have not yet been unveiled. To fill this gap, a novel bio-inspired modular control architecture that merges a recurrent cerebellar-like loop for adaptive control and a Smith predictor controller is proposed. The goal is to provide accurate anticipatory corrections to the generation of the motor commands in spite of sensory delays and to validate the robustness of the proposed control method to input and physical dynamic changes. The outcome of the proposed architecture with other two control schemes that do not include the Smith control strategy or the cerebellar-like corrections are compared. The results obtained on four sets of experiments confirm that the cerebellum-like circuit provides more effective corrections when only the Smith strategy is adopted and that minor tuning in the parameters, fast adaptation and reproducible configuration are enabled.

Nonlinear Bayesian inversion for estimating water pipeline dimensional and material parameters using acoustic wave dispersion

Monitoring pipeline thinning and material degeneration is becoming important for water-filled pipeline condition assessment. In this paper, an inverse method is proposed for estimating a pipeline's dimensional and material parameters using the dispersion characteristics of its modal wavenumbers. The inverse method is established by matching observed wavenumber dispersion characteristics of the water-filled pipeline with forward model predictions, where pipeline inner radius, thickness and density, and longitudinal and transverse wave speeds of the pipeline wall material are taken as unknown parameters. To account for the strong nonlinearity of the inverse problem and improve inversion efficiency, a Bayesian inversion scheme is formulated using a parallel-tempering Markov chain Monte Carlo approach. The characteristics and the performance of the proposed inverse method are investigated by systematic simulations which cover the impact of the number of modes utilized, dispersion frequency interval and observation errors. Laboratory experiments are utilized to validate the inversion method using wavenumber dispersion observations (below 50 kHz) from three metallic pipelines all with the same outer radius but different wall thicknesses and materials. The uncertainties of the estimated dimensional parameters are found to be lower than 0.2 mm and different materials are successfully distinguished and identified for the three pipelines.

Evaluation of High-Throughput Chemical Exposure Models: Case Study of Matched Environmental and Biological Measurements

U.S. EPA, under its ExpoCast program, is developing high-throughput near-field modeling methods to estimate chemical exposure and to provide real-world context to high-throughput screening (HTS) data. These novel modeling methods include reverse methods, which infer parent chemical exposures from biomonitoring measurements, and forward models to predict multi-pathway exposures from chemical use information and/or residential media concentrations. Both modeling methods were used to characterize the relationship between near-field environmental and biomarker measurements. Indoor air, house dust, and urine samples of 120 females (aged 60 to 80 years) were analyzed. In the measured data, 78% of the residential media measurements and 54% of the urine measurements were below the limit of quantification. Due to the degree of censoring in both air and dust data, we applied a latent variable estimation approach based on an expected relationship between air and dust concentrations. Using the partitioning model of Weschler and Nazaroff (Atmos. Env., 2010) and assumptions about the variance in concentrations within homes, we jointly estimated chemical-specific geometric mean air and dust concentrations from the observed measurements. This resulted in 14 chemicals with matched air, dust, and urine metabolite data. The indoor air and dust concentrations were compared to population median exposures inferred from urine metabolite concentrations using a reverse-dosimetry approach. Median air and dust concentrations were found to be correlated with inferred exposures; forward model predictions were used to characterize pathway contributions to aggregate exposures. These results demonstrate that the forward and reverse methods being developed in ExpoCast can predict intake amounts and routes of exposure, and that these models can also identify exposure pathways that contribute to biomarker concentrations in the general population. This abstract does not reflect EPA policy.

A vector radiative transfer model for sea-surface salinity retrieval from space: a non-raining case

ABSTRACTSea-surface salinity (SSS) can be measured from space using a microwave sensor. However, achieving the desired accuracy in SSS retrieval is challenging due to the lower sensitivity of the brightness temperature to SSS especially at low sea-surface temperature conditions. The retrieval accuracy can be further degraded due to the atmospheric and sea-surface effects (including emission and reflection), which require more accurate correction methods based on the radiative transfer model. In this article, a vector radiative transfer model (VRTM) was developed based on a matrix operator method that considers the ocean–atmosphere system under non-raining conditions. The results from this model were compared with measurement data provided by the Soil Moisture and Ocean Salinity (SMOS) satellite sensor and the results from two other RT models (RT4 model and a forward model of the European Space Agency, ESA). Statistical evaluation of these results revealed that estimation errors of top of atmosphere (TOA) radiance by the VRTM model was less than 0.3% as compared to the RT4 model results. The difference of the brightness temperatures predicted by the VRTM model and measured by the SMOS was within 1.5 K which was better than the ESA’s forward model predictions. These results suggest that the VRTM is relatively more accurate and has high computational efficiency for simulating the TOA brightness temperature for various scientific research and remote-sensing applications.

Self-monitoring in the cerebral cortex: Neural responses to small pitch shifts in auditory feedback during speech production

Speaking is a complex motor skill which requires near instantaneous integration of sensory and motor-related information. Current theory hypothesizes a complex interplay between motor and auditory processes during speech production, involving the online comparison of the speech output with an internally generated forward model. To examine the neural correlates of this intricate interplay between sensory and motor processes, the current study uses altered auditory feedback (AAF) in combination with magnetoencephalography (MEG). Participants vocalized the vowel/e/and heard auditory feedback that was temporarily pitch-shifted by only 25 cents, while neural activity was recorded with MEG. As a control condition, participants also heard the recordings of the same auditory feedback that they heard in the first half of the experiment, now without vocalizing. The participants were not aware of any perturbation of the auditory feedback. We found auditory cortical areas responded more strongly to the pitch shifts during vocalization. In addition, auditory feedback perturbation resulted in spectral power increases in the θ and lower β bands, predominantly in sensorimotor areas. These results are in line with current models of speech production, suggesting auditory cortical areas are involved in an active comparison between a forward model's prediction and the actual sensory input. Subsequently, these areas interact with motor areas to generate a motor response. Furthermore, the results suggest that θ and β power increases support auditory-motor interaction, motor error detection and/or sensory prediction processing.

Spatial-aware source estimation in building downwash environments

This paper introduces a concept called ”spatial-aware” source-term estimation (STE), which utilizes the acquired knowledge of how the forward model (FM) predictions compare against observations spatially to enhance the overall quality of source estimation. We evaluated the effectiveness of the proposed method against a wind tunnel experiment simulating low-stack plume dispersion in a building downwash environment with known source locations and varying building aspect ratios and building angles relative to wind direction, selected to achieve the balance of physical complexity and environmental controllability. Specifically, we adopted AERMOD as the FM, constructed spatially-resolved error models by comparing the FM predictions against measured concentrations for one configuration, referred to as the training case. Then the error models were applied to the rest of the configurations, referred to as test cases, for estimating emission rates. The results show that the spatial-aware STE can improve the estimation results, due to closer agreement between the error models and error distribution in terms of shapes and/or magnitudes. For example, using the vertical measurements around the H/W = 1/2 building with wind direction perpendicular to the building as the training case and applying the spatially-resolved error models, the estimated emission rates for all test cases with same building aspect ratio differ from the true emission rate by less than 30%, compared to 63.2%–236% without applying any error models. We argue that parallel efforts, i.e., improving the accuracy of FMs and quantifying the FM performance are needed to further advance the science and applications of source estimation.

Foot placement relies on state estimation during visually guided walking.

Accurate foot placement is essential for safe walking. We used computational models and human walking experiments to test how our nervous system achieves this accuracy. We find that our control of foot placement beneficially combines sensory feedback with internal forward model predictions to accurately estimate the body's state. Our results match recent computational neuroscience findings for reaching movements, suggesting that state estimation is a general mechanism of human motor control.

Automatic calibration of stratigraphic forward models for predicting reservoir presence in exploration

Understanding and predicting reservoir presence and characteristics at regional to basin scales is important for evaluating risk and uncertainty in hydrocarbon exploration. Simulating reservoir distribution within a basin by a stratigraphic forward model enables the integration of available prior information with fundamental geologic processes embedded in the numerical model. Stratigraphic forward model predictions can be significantly improved by calibrating the models to independent constraints, such as thicknesses from seismic or well data. A three-dimensional basin-scale stratigraphic forward-modeling tool is coupled with an inversion algorithm. The inversion algorithm is a modification of the neighborhood algorithm (a type of genetic algorithm), which is designed to sample complex multimodal objective functions and is parallelized on computer clusters to accelerate convergence. The process generates a set of representative geological models that are consistent with prior ranges for uncertain parameters, calibration constraints, and associated tolerance thresholds. The workflow is first demonstrated on two data sets: a synthetic example based on a clastic passive margin and a real hydrocarbon exploration example for slope and basin-floor stratigraphic traps in the Neocomian (Lower Cretaceous) of the West Siberian Basin. The analysis of calibrated models provides constraints on stratigraphic controls, and allows prediction of locations with higher potential to develop stratigraphic traps. These locations are related to complex interactions between paleobathymetry, subsidence, eustatic fluctuations, characteristics of sediment-input sources, and sediment-transport parameters. Results show the potential of stratigraphic forward modeling combined with inverse methods as an additional tool to support conventional play-based exploration and reservoir-presence prediction.

Using geodetic GPS receivers to measure vegetation water content

A GPS-based method is presented to measure vegetation water content. Commercially available geodetic-quality GPS receivers and antennas are used. The method is tested using GPS measurements collected over three field seasons. The GPS data are compared with in situ data for three plant types: desert grass, wheat, and alfalfa. The GPS retrievals of vegetation water content are based on the GPS signal-to-noise ratio (SNR) data. Instrumental issues that affect the SNR data are discussed, particularly satellite transmit power variations, footprint variations, and temperature effects. The amplitudes of the SNR data show a nearly linear relationship to the water content in grasses (0---0.5 kg/m2) and wheat crops (0---0.9 kg/m2). As predicted by theory, this simple linear relationship breaks down in vegetation with heavy water content, such as alfalfa. The field results are consistent with forward model predictions, whose effect restricts the use of this simple linear model for vegetation to water content of less than ~1 kg/m2.

Accounting for prediction uncertainty when inferring subsurface fault slip

This study lays the groundwork for a new generation of earthquake source models based on a general formalism that rigorously quantifies and incorporates the impact of uncertainties in fault slip inverse problems. We distinguish two sources of uncertainty when considering the discrepancy between data and forward model predictions. The first class of error is induced by imperfect measurements and is often referred to as observational error. The second source of uncertainty is generally neglected and corresponds to the prediction error, that is the uncertainty due to imperfect forward modelling. Yet the prediction error can be shown to scale approximately with the size of earthquakes and thus can dwarf the observational error, particularly for large events. Both sources of uncertainty can be formulated using the misfit covariance matrix, C_χ, which combines a covariance matrix for observation errors, C_d and a covariance matrix for prediction errors, C_p, associated with inaccurate model predictions. We develop a physically based stochastic forward model to treat the model prediction uncertainty and show how C_p can be constructed to explicitly account for some of the inaccuracies in the earth model. Based on a first-order perturbation approach, our formalism relates C_p to uncertainties on the elastic parameters of different regions (e.g. crust, mantle, etc.). We demonstrate the importance of including C_p using a simple example of an infinite strike-slip fault in the quasi-static approximation. In this toy model, we treat only uncertainties in the 1-D depth distribution of the shear modulus. We discuss how this can be extended to general 3-D cases and applied to other parameters (e.g. fault geometry) using our formalism for C_p. The improved modelling of C_p is expected to lead to more reliable images of the earthquake rupture, that are more resistant to overfitting of data and include more realistic estimates of uncertainty on inferred model parameters.

Prediction at all levels: forward model predictions can enhance comprehension

We discuss two limitations of Hickok's account. First, we propose that ideas from motor control and planning should be brought wholesale into psycholinguistics so that processing at every level of the linguistic hierarchy (from concepts to sounds) should be recast in terms of forward model predictions and implementation. Second, we argue that motor involvement can sometimes enhance perception. We conclude that our account is consistent with a dual route model of comprehension in which different routes to prediction can interact.

Antarctic contribution to sea level rise observed by GRACE with improved GIA correction

Antarctic volume changes during the past 21 thousand years are smaller than previously thought, and here we construct an ice sheet history that drives a forward model prediction of the glacial isostatic adjustment (GIA) gravity signal. The new model, in turn, should give predictions that are constrained with recent uplift data. The impact of the GIA signal on a Gravity Recovery and Climate Experiment (GRACE) Antarctic mass balance estimate depends on the specific GRACE analysis method used. For the method described in this paper, the GIA contribution to the apparent surface mass change is re‐evaluated to be +55±13 Gt/yr by considering a revised ice history model and a parameter search for vertical motion predictions that best fit the GPS observations at 18 high‐quality stations. Although the GIA model spans a range of possible Earth rheological structure values, the data are not yet sufficient for solving for a preferred value of upper and lower mantle viscosity nor for a preferred lithospheric thickness. GRACE monthly solutions from the Center for Space Research Release 04 (CSR‐RL04) release time series from January 2003 to the beginning of January 2012, uncorrected for GIA, yield an ice mass rate of +2.9± 29 Gt/yr. The new GIA correction increases the solved‐for ice mass imbalance of Antarctica to −57±34 Gt/yr. The revised GIA correction is smaller than past GRACE estimates by about 50 to 90 Gt/yr. The new upper bound to the sea level rise from the Antarctic ice sheet, averaged over the time span 2003.0–2012.0, is about 0.16±0.09 mm/yr.

Time frequency analysis techniques for long range sediment tomography

Long range sediment tomography inversion technique requires accurate estimation of the arrival times of acoustic normal modes. The inversion technique is based on minimizing the difference between the forward model predictions and the data using a global optimization scheme. Predictions are computed using a trial parameter set which is iteratively modified until the algorithm converges. The modal arrival times are calculated from the time-frequency analysis of broadband acoustic data collected on a single hydrophone. During the initial stages of the development of the inversion scheme, Fourier-based spectrograms and wavelet-based scalograms were used. Taking advantage of some of the recent developments, dispersion based short time Fourier transform (DSTFT) and warping transform techniques were used for the time-frequency analysis, in recent times. This paper will summarize and compare the performance of these time frequency techniques in the context of long range sediment tomography. Data from some of the recent field tests (Shelfbreak Primer and Shallow Water-06 experiments) will be analyzed for this study. Finally, time-frequency analysis performance of another new technique, Modified S transform, will be examined using the field data. [Work supported by Office of Naval Research].

Alteration in forward model prediction of sensory outcome of motor action in focal hand dystonia

Focal hand dystonia in musicians is a movement disorder affecting highly trained movements. Rather than being a pure motor disorder related to movement execution only, movement planning, error prediction, and sensorimotor integration are also impaired. Internal models (IMs), of which two types, forward and inverse models have been described and most likely processed in the cerebellum, are known to be involved in these tasks. Recent results indicate that the cerebellum may be involved in the pathophysiology of focal dystonia (FD). Thus, the aim of our study was to investigate whether an IM deficit plays a role in FD. We focused on the forward model (FM), which predicts sensory consequences of motor commands and allows the discrimination between external sensory input and input deriving from motor action. We investigated 19 patients, aged 19–59 and 19 healthy musicians aged 19–36 as controls. Tactile stimuli were applied to fingers II–V of both hands by the experimenter or the patient. After each stimulus the participant rated the stimulus intensity on a scale between 0 (no sensation) and 1 (maximal intensity). The difference of perceived intensity between self- and externally applied (EA) stimuli was then calculated for each finger. For assessing differences between patients and controls we performed a cluster analysis of the affected hand and the corresponding hand of the controls using the fingers II–V as variables in a 4-dimensional hyperspace (chance level = 0.5). Using a cluster analysis, we found a correct classification of the affected finger in 78.9–94.7%. There was no difference between patients and healthy controls of the absolute value of the perceived stimulus intensity. Our results suggest an altered FM function in focal hand dystonia. It has the potential of suggesting a neural correlate within the cerebellum and of helping integrate findings with regard to altered sensorimotor processing and altered prediction in FD in a single framework.

Prediction and imitation in speech

It has been suggested that intra- and inter-speaker variability in speech are correlated. Interlocutors have been shown to converge on various phonetic dimensions. In addition, speakers imitate the phonetic properties of voices they are exposed to in shadowing, repetition, and even passive listening tasks. We review three theoretical accounts of speech imitation and convergence phenomena: (i) the Episodic Theory (ET) of speech perception and production (Goldinger, 1998); (ii) the Motor Theory (MT) of speech perception (Liberman and Whalen, 2000; Galantucci et al., 2006); (iii) Communication Accommodation Theory (CAT; Giles and Coupland, 1991; Giles et al., 1991). We argue that no account is able to explain all the available evidence. In particular, there is a need to integrate low-level, mechanistic accounts (like ET and MT), and higher-level accounts (like CAT). We propose that this is possible within the framework of an integrated theory of production and comprehension (Pickering and Garrod, 2013). Similarly to both ET and MT, this theory assumes parity between production and perception. Uniquely, however, it posits that listeners simulate speakers' utterances by computing forward-model predictions at many different levels, which are then compared to the incoming phonetic input. In our account phonetic imitation can be achieved via the same mechanism that is responsible for sensorimotor adaptation; i.e., the correction of prediction errors. In addition, the model assumes that the degree to which sensory prediction errors lead to motor adjustments is context-dependent. The notion of context subsumes both the preceding linguistic input and non-linguistic attributes of the situation (e.g., the speaker's and listener's social identities, their conversational roles, the listener's intention to imitate).

Wheat Canopy Structure and Surface Roughness Effects on Multiangle Observations at L-Band

The multiangle observation capability of the Soil Moisture and Ocean Salinity mission is expected to significantly improve the inversion of soil microwave emissions for soil moisture, by enabling the simultaneous retrieval of the vegetation optical depth and other surface parameters. Consequently, this paper investigates the relationship between soil moisture and brightness temperature at multiple incidence angles using airborne L-band data from the National Airborne Field Experiment in Australia in 2005. A forward radio brightness model was used to predict the passive microwave response at a range of incidence angles, given the following inputs: 1) ground-measured soil and vegetation properties and 2) default model parameters for vegetation and roughness characterization. Simulations were made across various dates and locations with wheat cover and evaluated against the available airborne observations. The comparison showed a significant underestimation of the measured brightness temperatures by the model. This discrepancy subsequently led to soil moisture retrieval errors of up to 0.3 m3/m3. Further analysis found the following: 1) The roughness value HR was too low, which was then adjusted as a function of the soil moisture, and 2) the vegetation structure parameters tth and ttv required optimization, yielding new values of tth = 0.2 and ttv = 1.4 from calibration to a single flight. Testing the optimized parameterization for different moisture conditions and locations found that the root-mean-square simulation error between the forward model predictions and the airborne observations was improved from 31.3 K (26.5 K) to 2.3 K (5.3 K) for wet (dry) soil moisture condition.