scholarly journals Remembrance of Inferences Past

2017 ◽  
Author(s):  
Ishita Dasgupta ◽  
Eric Schulz ◽  
Noah D. Goodman ◽  
Samuel J. Gershman
Keyword(s):  
Probabilistic Reasoning ◽  
Probability Distributions ◽  
Computational Cost ◽  
Limited Resources ◽  
Cognitive Resources ◽  
Natural Scenes ◽  
Computationally Efficient ◽  
Bayesian Models Of Cognition ◽  
The Brain ◽  
Entire Distribution

AbstractBayesian models of cognition assume that people compute probability distributions over hypotheses. However, the required computations are frequently intractable or prohibitively expensive. Since people often encounter many closely related distributions, selective reuse of computations (amortized inference) is a computationally efficient use of the brain’s limited resources. We present three experiments that provide evidence for amortization in human probabilistic reasoning. When sequentially answering two related queries about natural scenes, participants’ responses to the second query systematically depend on the structure of the first query. This influence is sensitive to the content of the queries, only appearing when the queries are related. Using a cognitive load manipulation, we find evidence that people amortize summary statistics of previous inferences, rather than storing the entire distribution. These findings support the view that the brain trades off accuracy and computational cost, to make efficient use of its limited cognitive resources to approximate probabilistic inference.

scholarly journals Amortized Hypothesis Generation

2017 ◽  
Author(s):  
Ishita Dasgupta ◽  
Eric Schulz ◽  
Noah D. Goodman ◽  
Samuel J. Gershman
Keyword(s):  
Cognitive Load ◽  
Probability Distributions ◽  
Hypothesis Generation ◽  
Natural Scenes ◽  
Summary Statistics ◽  
Summary Statistic ◽  
Computationally Efficient ◽  
Present Evidence ◽  
Bayesian Models Of Cognition ◽  
The Brain

AbstractBayesian models of cognition posit that people compute probability distributions over hypotheses, possibly by constructing a sample-based approximation. Since people encounter many closely related distributions, a computationally efficient strategy is to selectively reuse computations – either the samples themselves or some summary statistic. We refer to these reuse strategies as amortized inference. In two experiments, we present evidence consistent with amortization. When sequentially answering two related queries about natural scenes, we show that answers to the second query vary systematically depending on the structure of the first query. Using a cognitive load manipulation, we find evidence that people cache summary statistics rather than raw sample sets. These results enrich our notions of how the brain approximates probabilistic inference.

scholarly journals A theory of learning to infer

2019 ◽  
Author(s):  
Ishita Dasgupta ◽  
Eric Schulz ◽  
Joshua B. Tenenbaum ◽  
Samuel J. Gershman
Keyword(s):  
Probabilistic Reasoning ◽  
Probability Distributions ◽  
General Purpose ◽  
Response Variability ◽  
Recognition Model ◽  
Prior Probabilities ◽  
Base Rate Neglect ◽  
Low Probability ◽  
Computational Resources ◽  
The Brain

AbstractBayesian theories of cognition assume that people can integrate probabilities rationally. However, several empirical findings contradict this proposition: human probabilistic inferences are prone to systematic deviations from optimality. Puzzlingly, these deviations sometimes go in opposite directions. Whereas some studies suggest that people under-react to prior probabilities (base rate neglect), other studies find that people under-react to the likelihood of the data (conservatism). We argue that these deviations arise because the human brain does not rely solely on a general-purpose mechanism for approximating Bayesian inference that is invariant across queries. Instead, the brain is equipped with a recognition model that maps queries to probability distributions. The parameters of this recognition model are optimized to get the output as close as possible, on average, to the true posterior. Because of our limited computational resources, the recognition model will allocate its resources so as to be more accurate for high probability queries than for low probability queries. By adapting to the query distribution, the recognition model “learns to infer.” We show that this theory can explain why and when people under-react to the data or the prior, and a new experiment demonstrates that these two forms of under-reaction can be systematically controlled by manipulating the query distribution. The theory also explains a range of related phenomena: memory effects, belief bias, and the structure of response variability in probabilistic reasoning. We also discuss how the theory can be integrated with prior sampling-based accounts of approximate inference.

scholarly journals Compressed pseudo-SLAM: pseudorange-integrated compressed simultaneous localisation and mapping for unmanned aerial vehicle navigation

2021 ◽  
pp. 1-13
Author(s):  
Jonghyuk Kim ◽  
Jose Guivant ◽  
Martin L. Sollie ◽  
Torleiv H. Bryne ◽  
Tor Arne Johansen
Keyword(s):  
Unmanned Aerial Vehicle ◽  
Computational Cost ◽  
Satellite System ◽  
Measurement Unit ◽  
Electronic Systems ◽  
Computationally Efficient ◽  
Global Correlation ◽  
Aerial Vehicle ◽  
Correlation Information ◽  
Global Navigation Satellite

Abstract This paper addresses the fusion of the pseudorange/pseudorange rate observations from the global navigation satellite system and the inertial–visual simultaneous localisation and mapping (SLAM) to achieve reliable navigation of unmanned aerial vehicles. This work extends the previous work on a simulation-based study [Kim et al. (2017). Compressed fusion of GNSS and inertial navigation with simultaneous localisation and mapping. IEEE Aerospace and Electronic Systems Magazine, 32(8), 22–36] to a real-flight dataset collected from a fixed-wing unmanned aerial vehicle platform. The dataset consists of measurements from visual landmarks, an inertial measurement unit, and pseudorange and pseudorange rates. We propose a novel all-source navigation filter, termed a compressed pseudo-SLAM, which can seamlessly integrate all available information in a computationally efficient way. In this framework, a local map is dynamically defined around the vehicle, updating the vehicle and local landmark states within the region. A global map includes the rest of the landmarks and is updated at a much lower rate by accumulating (or compressing) the local-to-global correlation information within the filter. It will show that the horizontal navigation error is effectively constrained with one satellite vehicle and one landmark observation. The computational cost will be analysed, demonstrating the efficiency of the method.

Two-dimensional Bayesian inversion of magnetotelluric data using trans-dimensional Gaussian processes

2021 ◽  
Author(s):  
Daniel Blatter ◽  
Anandaroop Ray ◽  
Kerry Key
Keyword(s):  
Gaussian Process ◽  
Process Model ◽  
Field Data ◽  
Probability Distributions ◽  
Bulk Composition ◽  
Computational Cost ◽  
Model Space ◽  
Bayesian Inversion ◽  
Adaptive Finite Element ◽  
Magnetotelluric Data

Summary Bayesian inversion of electromagnetic data produces crucial uncertainty information on inferred subsurface resistivity. Due to their high computational cost, however, Bayesian inverse methods have largely been restricted to computationally expedient 1D resistivity models. In this study, we successfully demonstrate, for the first time, a fully 2D, trans-dimensional Bayesian inversion of magnetotelluric data. We render this problem tractable from a computational standpoint by using a stochastic interpolation algorithm known as a Gaussian process to achieve a parsimonious parametrization of the model vis-a-vis the dense parameter grids used in numerical forward modeling codes. The Gaussian process links a trans-dimensional, parallel tempered Markov chain Monte Carlo sampler, which explores the parsimonious model space, to MARE2DEM, an adaptive finite element forward solver. MARE2DEM computes the model response using a dense parameter mesh with resistivity assigned via the Gaussian process model. We demonstrate the new trans-dimensional Gaussian process sampler by inverting both synthetic and field magnetotelluric data for 2D models of electrical resistivity, with the field data example converging within 10 days on 148 cores, a non-negligible but tractable computational cost. For a field data inversion, our algorithm achieves a parameter reduction of over 32x compared to the fixed parameter grid used for the MARE2DEM regularized inversion. Resistivity probability distributions computed from the ensemble of models produced by the inversion yield credible intervals and interquartile plots that quantitatively show the non-linear 2D uncertainty in model structure. This uncertainty could then be propagated to other physical properties that impact resistivity including bulk composition, porosity and pore-fluid content.

Hybrid LES Approach for Practical Turbomachinery Flows—Part I: Hierarchy and Example Simulations

2011 ◽  
Vol 134 (2) ◽  
Author(s):  
Paul Tucker ◽  
Simon Eastwood ◽  
Christian Klostermeier ◽  
Richard Jefferson-Loveday ◽  
James Tyacke ◽  
...  
Keyword(s):  
Convex Surface ◽  
Computational Cost ◽  
Companion Paper ◽  
Navier Stokes ◽  
Computationally Efficient ◽  
Turbulent Structures ◽  
Hybrid Strategy ◽  
Eddy Simulation ◽  
Related Approach ◽  
Les Model

Unlike Reynolds-averaged Navier–Stokes (RANS) models that need calibration for different flow classes, LES (where larger turbulent structures are resolved by the grid and smaller modeled in a fashion reminiscent of RANS) offers the opportunity to resolve geometry dependent turbulence as found in complex internal flows—albeit at substantially higher computational cost. Based on the results for a broad range of studies involving different numerical schemes, large eddy simulation (LES) models and grid topologies, an LES hierarchy and hybrid LES related approach is proposed. With the latter, away from walls, no LES model is used, giving what can be termed numerical LES (NLES). This is relatively computationally efficient and makes use of the dissipation present in practical industrial computational fluid dynamics (CFD) programs. Near walls, RANS modeling is used to cover over numerous small structures, the LES resolution of which is generally intractable with current computational power. The linking of the RANS and NLES zones through a Hamilton–Jacobi equation is advocated. The RANS-NLES hybridization makes further sense for compressible flow solvers, where, as the Mach number tends to zero at walls, excessive dissipation can occur. The hybrid strategy is used to predict flow over a rib roughened surface and a jet impinging on a convex surface. These cases are important for blade cooling and show encouraging results. Further results are presented in a companion paper.

An Efficient Preconditioner for Linear System Solution in Multi-Domain Modeling of the Circulatory System

2013 ◽  
Author(s):  
Mahdi Esmaily Moghadam ◽  
Yuri Bazilevs ◽  
Tain-Yen Hsia ◽  
Alison Marsden
Keyword(s):  
Strong Coupling ◽  
Large Scale ◽  
Computational Cost ◽  
Circulatory System ◽  
Flow Simulation ◽  
Global Dynamics ◽  
Domain Modeling ◽  
Computationally Efficient ◽  
Lumped Parameter ◽  
System Solution

A closed-loop lumped parameter network (LPN) coupled to a 3D domain is a powerful tool that can be used to model the global dynamics of the circulatory system. Coupling a 0D LPN to a 3D CFD domain is a numerically challenging problem, often associated with instabilities, extra computational cost, and loss of modularity. A computationally efficient finite element framework has been recently proposed that achieves numerical stability without sacrificing modularity [1]. This type of coupling introduces new challenges in the linear algebraic equation solver (LS), producing an strong coupling between flow and pressure that leads to an ill-conditioned tangent matrix. In this paper we exploit this strong coupling to obtain a novel and efficient algorithm for the linear solver (LS). We illustrate the efficiency of this method on several large-scale cardiovascular blood flow simulation problems.

scholarly journals Recurrent Multi-Fiber Network for 3D MRI Brain Tumor Segmentation

Symmetry ◽  
2021 ◽  
Vol 13 (2) ◽  
pp. 320
Author(s):  
Yue Zhao ◽  
Xiaoqiang Ren ◽  
Kun Hou ◽  
Wentao Li
Keyword(s):  
Brain Tumor ◽  
Network Architecture ◽  
Contextual Information ◽  
Computational Cost ◽  
Disease Diagnosis ◽  
Tumor Segmentation ◽  
Brain Tumor Segmentation ◽  
Fiber Network ◽  
3D Network ◽  
The Brain

Automated brain tumor segmentation based on 3D magnetic resonance imaging (MRI) is critical to disease diagnosis. Moreover, robust and accurate achieving automatic extraction of brain tumor is a big challenge because of the inherent heterogeneity of the tumor structure. In this paper, we present an efficient semantic segmentation 3D recurrent multi-fiber network (RMFNet), which is based on encoder–decoder architecture to segment the brain tumor accurately. 3D RMFNet is applied in our paper to solve the problem of brain tumor segmentation, including a 3D recurrent unit and 3D multi-fiber unit. First of all, we propose that recurrent units segment brain tumors by connecting recurrent units and convolutional layers. This quality enhances the model’s ability to integrate contextual information and is of great significance to enhance the contextual information. Then, a 3D multi-fiber unit is added to the overall network to solve the high computational cost caused by the use of a 3D network architecture to capture local features. 3D RMFNet combines both advantages from a 3D recurrent unit and 3D multi-fiber unit. Extensive experiments on the Brain Tumor Segmentation (BraTS) 2018 challenge dataset show that our RMFNet remarkably outperforms state-of-the-art methods, and achieves average Dice scores of 89.62%, 83.65% and 78.72% for the whole tumor, tumor core and enhancing tumor, respectively. The experimental results prove our architecture to be an efficient tool for brain tumor segmentation accurately.

A Computationally Efficient Predictive Controller for Lane Keeping of Semi-Autonomous Vehicles

2014 ◽  
Author(s):  
Changchun Liu ◽  
Chankyu Lee ◽  
Andreas Hansen ◽  
J. Karl Hedrick ◽  
Jieyun Ding
Keyword(s):  
Autonomous Vehicles ◽  
Piecewise Linear ◽  
Computational Cost ◽  
Parametric Programming ◽  
Linear Feedback ◽  
Loop Model ◽  
Computationally Efficient ◽  
Feedback Function ◽  
Predictive Controller ◽  
Lane Keeping

Model predictive control (MPC) is a popular technique for the development of active safety systems. However, its high computational cost prevents it from being implemented on lower-cost hardware. This paper presents a computationally efficient predictive controller for lane keeping assistance systems. The controller shares control with the driver, and applies a correction steering when there is a potential lane departure. Using the explicit feedback MPC, a multi-parametric nonlinear programming problem with a human-in-the-loop model and safety constraints is formulated. The cost function is chosen as the difference between the linear state feedback function to be determined and the resultant optimal control sequence of the MPC problem solved off-line given the current state. The piecewise linear feedback function is obtained by solving the parametric programming with an approximation approach. The effectiveness of the controller is evaluated through numerical simulations.

scholarly journals Intrinsic spatial knowledge about terrestrial ecology favors the tall for judging distance

2016 ◽  
Vol 2 (8) ◽  
pp. e1501070 ◽  
Author(s):  
Liu Zhou ◽  
Teng Leng Ooi ◽  
Zijiang J. He
Keyword(s):  
Target Location ◽  
Spatial Relationship ◽  
Ground Surface ◽  
Spatial Knowledge ◽  
Superior Performance ◽  
Natural Scenes ◽  
Terrestrial Environment ◽  
Optical Images ◽  
Intermediate Distance ◽  
The Brain

Our sense of vision reliably directs and guides our everyday actions, such as reaching and walking. This ability is especially fascinating because the optical images of natural scenes that project into our eyes are insufficient to adequately form a perceptual space. It has been proposed that the brain makes up for this inadequacy by using its intrinsic spatial knowledge. However, it is unclear what constitutes intrinsic spatial knowledge and how it is acquired. We investigated this question and showed evidence of an ecological basis, which uses the statistical spatial relationship between the observer and the terrestrial environment, namely, the ground surface. We found that in dark and reduced-cue environments where intrinsic knowledge has a greater contribution, perceived target location is more accurate when referenced to the ground than to the ceiling. Furthermore, taller observers more accurately localized the target. Superior performance was also observed in the full-cue environment, even when we compensated for the observers’ heights by having the taller observer sit on a chair and the shorter observers stand on a box. Although fascinating, this finding dovetails with the prediction of the ecological hypothesis for intrinsic spatial knowledge. It suggests that an individual’s accumulated lifetime experiences of being tall and his or her constant interactions with ground-based objects not only determine intrinsic spatial knowledge but also endow him or her with an advantage in spatial ability in the intermediate distance range.

scholarly journals Task-Dependent Warping of Semantic Representations During Search for Visual Action Categories

2021 ◽  
Author(s):  
Mo Shahdloo ◽  
Emin Çelik ◽  
Burcu A Urgen ◽  
Jack L. Gallant ◽  
Tolga Çukur
Keyword(s):  
Visual Search ◽  
Social Interactions ◽  
Brain Activity ◽  
Semantic Representation ◽  
Natural Scenes ◽  
Action Perception ◽  
Semantic Representations ◽  
Object Categories ◽  
Search For Objects ◽  
The Brain

Object and action perception in cluttered dynamic natural scenes relies on efficient allocation of limited brain resources to prioritize the attended targets over distractors. It has been suggested that during visual search for objects, distributed semantic representation of hundreds of object categories is warped to expand the representation of targets. Yet, little is known about whether and where in the brain visual search for action categories modulates semantic representations. To address this fundamental question, we studied human brain activity recorded via functional magnetic resonance imaging while subjects viewed natural movies and searched for either communication or locomotion actions. We find that attention directed to action categories elicits tuning shifts that warp semantic representations broadly across neocortex, and that these shifts interact with intrinsic selectivity of cortical voxels for target actions. These results suggest that attention serves to facilitate task performance during social interactions by dynamically shifting semantic selectivity towards target actions, and that tuning shifts are a general feature of conceptual representations in the brain.

Sign in / Sign up

Export Citation Format

Share Document