scholarly journals Statistical regression for efficient high-dimensional modeling of analog and mixed-signal performance variations

Author(s):  
Xin Li ◽  
Hongzhou Liu
2021 ◽  
Author(s):  
◽  
John Lewis

<p>Movie and game production is very laborious, frequently involving hundreds of person-years for a single project. At present this work is difficult to fully automate, since it involves subjective and artistic judgments.  Broadly speaking, in this thesis we explore an approach that works with the artist, accelerating their work without attempting to replace them. More specifically, we describe an “example-based” approach, in which artists provide examples of the desired shapes of the character, and the results gradually improve as more examples are given. Since a character’s skin shape deforms as the pose or expression changes, or particular problem will be termed character deformation.  The overall goal of this thesis is to contribute a complete investigation and development of an example-based approach to character deformation. A central observation guiding this research is that character animation can be formulated as a high-dimensional problem, rather than the two- or three-dimensional viewpoint that is commonly adopted in computer graphics. A second observation guiding our inquiry is that statistical learning concepts are relevant. We show that example-based character animation algorithms can be informed, developed, and improved using these observations.  This thesis provides definitive surveys of example-based facial and body skin deformation.  This thesis analyzes the two leading families of example-based character deformation algorithms from the point of view of statistical regression. In doing so we show that a wide variety of existing tools in machine learning are applicable to our problem. We also identify several techniques that are not suitable due to the nature of the training data, and the high-dimensional nature of this regression problem. We evaluate the design decisions underlying these example-based algorithms, thus providing the groundwork for a ”best practice” choice of specific algorithms.  This thesis develops several new algorithms for accelerating example-based facial animation. The first algorithm allows unspecified degrees of freedom to be automatically determined based on the style of previous, completed animations. A second algorithm allows rapid editing and control of the process of transferring motion capture of a human actor to a computer graphics character.  The thesis identifies and develops several unpublished relations between the underlying mathematical techniques.  Lastly, the thesis provides novel tutorial derivations of several mathematical concepts, using only the linear algebra tools that are likely to be familiar to experts in computer graphics.  Portions of the research in this thesis have been published in eight papers, with two appearing in premier forums in the field.</p>


2021 ◽  
Author(s):  
Taylor W Webb ◽  
Kiyofumi Miyoshi ◽  
Tsz Yan So ◽  
Sivananda Rajananda ◽  
Hakwan Lau

Previous work has sought to understand decision confidence as a prediction of the probability that a decision will be correct, leading to debate over whether these predictions are optimal, and whether they rely on the same decision variable as decisions themselves. This work has generally relied on idealized, low-dimensional modeling frameworks, such as signal detection theory or Bayesian inference, leaving open the question of how decision confidence operates in the domain of high-dimensional, naturalistic stimuli. To address this, we developed a deep neural network model optimized to assess decision confidence directly given high-dimensional inputs such as images. The model naturally accounts for a number of puzzling dissociations between decisions and confidence, suggests a principled explanation of these dissociations in terms of optimization for the statistics of sensory inputs, and makes the surprising prediction that, despite these dissociations, decisions and confidence depend on a common decision variable.


2020 ◽  
Vol 15 (3) ◽  
pp. 909-935 ◽  
Author(s):  
Xinming Yang ◽  
Naveen N. Narisetty

2021 ◽  
Author(s):  
◽  
John Lewis

<p>Movie and game production is very laborious, frequently involving hundreds of person-years for a single project. At present this work is difficult to fully automate, since it involves subjective and artistic judgments.  Broadly speaking, in this thesis we explore an approach that works with the artist, accelerating their work without attempting to replace them. More specifically, we describe an “example-based” approach, in which artists provide examples of the desired shapes of the character, and the results gradually improve as more examples are given. Since a character’s skin shape deforms as the pose or expression changes, or particular problem will be termed character deformation.  The overall goal of this thesis is to contribute a complete investigation and development of an example-based approach to character deformation. A central observation guiding this research is that character animation can be formulated as a high-dimensional problem, rather than the two- or three-dimensional viewpoint that is commonly adopted in computer graphics. A second observation guiding our inquiry is that statistical learning concepts are relevant. We show that example-based character animation algorithms can be informed, developed, and improved using these observations.  This thesis provides definitive surveys of example-based facial and body skin deformation.  This thesis analyzes the two leading families of example-based character deformation algorithms from the point of view of statistical regression. In doing so we show that a wide variety of existing tools in machine learning are applicable to our problem. We also identify several techniques that are not suitable due to the nature of the training data, and the high-dimensional nature of this regression problem. We evaluate the design decisions underlying these example-based algorithms, thus providing the groundwork for a ”best practice” choice of specific algorithms.  This thesis develops several new algorithms for accelerating example-based facial animation. The first algorithm allows unspecified degrees of freedom to be automatically determined based on the style of previous, completed animations. A second algorithm allows rapid editing and control of the process of transferring motion capture of a human actor to a computer graphics character.  The thesis identifies and develops several unpublished relations between the underlying mathematical techniques.  Lastly, the thesis provides novel tutorial derivations of several mathematical concepts, using only the linear algebra tools that are likely to be familiar to experts in computer graphics.  Portions of the research in this thesis have been published in eight papers, with two appearing in premier forums in the field.</p>


2017 ◽  
Vol 43 (1) ◽  
pp. 3-31 ◽  
Author(s):  
Adam C. Sales ◽  
Ben B. Hansen ◽  
Brian Rowan

In causal matching designs, some control subjects are often left unmatched, and some covariates are often left unmodeled. This article introduces “rebar,” a method using high-dimensional modeling to incorporate these commonly discarded data without sacrificing the integrity of the matching design. After constructing a match, a researcher uses the unmatched control subjects—the remnant—to fit a machine learning model predicting control potential outcomes as a function of the full covariate matrix. The resulting predictions in the matched set are used to adjust the causal estimate to reduce confounding bias. We present theoretical results to justify the method’s bias-reducing properties as well as a simulation study that demonstrates them. Additionally, we illustrate the method in an evaluation of a school-level comprehensive educational reform program in Arizona.


Sign in / Sign up

Export Citation Format

Share Document