國立虎尾科技大學 |

Statistical Analysis of Scientific Machine Learning.

紀錄類型:	書目-語言資料,手稿 : Monograph/item
正題名/作者:	Statistical Analysis of Scientific Machine Learning./
作者:	Lu, Yiping.
面頁冊數:	1 online resource (158 pages)
附註:	Source: Dissertations Abstracts International, Volume: 85-06, Section: B.
Contained By:	Dissertations Abstracts International85-06B.
標題:	Neural networks. -
電子資源:	click for full text (PQDT)
ISBN:	9798381019001

Statistical Analysis of Scientific Machine Learning.
Lu, Yiping.

Statistical Analysis of Scientific Machine Learning. - 1 online resource (158 pages)

Source: Dissertations Abstracts International, Volume: 85-06, Section: B.

Thesis (Ph.D.)--Stanford University, 2023.

Includes bibliographical references

Massive data collection and computational capabilities have enabled data-driven scientific discoveries and control of engineering systems. However, there are still several questions that should be answered to understand the fundamental limits of just how much can be discovered with data and what is the value of additional information. For example, 1) How can we learn a physics law or economic principle purely from data? 2) How hard is this task, both computationally and statistically? 3) What's the impact on hardness when we add further information (e.g., adding data, model information)? I'll answer these three questions in this thesis in two learning tasks. A key insight in both two cases is that using direct plug-in estimators can result in statistically suboptimal inference.For the first learning task, the thesis focus on variational formulations for differential equation models. I discuss a prototypical Poisson equation. I provide a minimax lower bound for this problem. Based on the lower bounds, I discover that the variance in the direct plug-in estimator makes sample complexity suboptimal. I also consider the optimization dynamic for different variational forms. Finally, based on our theory, I explain an implicit acceleration of using a Sobolev norm as the objective function for training.The second learning task this thesis discuss is (linear) operator learning, which has wide applications in causal inference, time series modeling, and conditional probability learning. I build the first min-max lower bound for this problem. The min-max rate has a particular structure where the more challenging parts of the input and output spaces determine the hardness of learning a linear operator. Analysis also shows that an intuitive discretization of the infinite-dimensional operator could lead to a sub-optimal statistical learning rate. Then, I'll discuss how, by suitably trading-off bias and variance, I can construct an estimator with an optimal learning rate for learning a linear operator between infinite dimension spaces. I also illustrate how this theory can inspire a multilevel machine-learning algorithm of potential practical use.

Electronic reproduction.
Ann Arbor, Mich. :
ProQuest,
2024

Mode of access: World Wide Web

ISBN: 9798381019001Subjects--Topical Terms:

1011215
Neural networks.
Index Terms--Genre/Form:

554714
Electronic books.

Statistical Analysis of Scientific Machine Learning.
LDR:03374ntm a22003257 4500 001 1148743
005 20240930100134.5
006 m o d
007 cr bn ---uuuuu
008 250605s2023 xx obm 000 0 eng d
020 $a 9798381019001
035 $a (MiAaPQ)AAI30726802
035 $a (MiAaPQ)STANFORDpp190dc8926
035 $a AAI30726802
040 $a MiAaPQ $b eng $c MiAaPQ $d NTU
100 1 $a Lu, Yiping. $3 1474786
245 1 0 $a Statistical Analysis of Scientific Machine Learning.
264 0 $c 2023
300 $a 1 online resource (158 pages)
336 $a text $b txt $2 rdacontent
337 $a computer $b c $2 rdamedia
338 $a online resource $b cr $2 rdacarrier
500 $a Source: Dissertations Abstracts International, Volume: 85-06, Section: B.
500 $a Advisor: Ying, Lexing;Mancilla, Jose Blanchet.
502 $a Thesis (Ph.D.)--Stanford University, 2023.
504 $a Includes bibliographical references
520 $a Massive data collection and computational capabilities have enabled data-driven scientific discoveries and control of engineering systems. However, there are still several questions that should be answered to understand the fundamental limits of just how much can be discovered with data and what is the value of additional information. For example, 1) How can we learn a physics law or economic principle purely from data? 2) How hard is this task, both computationally and statistically? 3) What's the impact on hardness when we add further information (e.g., adding data, model information)? I'll answer these three questions in this thesis in two learning tasks. A key insight in both two cases is that using direct plug-in estimators can result in statistically suboptimal inference.For the first learning task, the thesis focus on variational formulations for differential equation models. I discuss a prototypical Poisson equation. I provide a minimax lower bound for this problem. Based on the lower bounds, I discover that the variance in the direct plug-in estimator makes sample complexity suboptimal. I also consider the optimization dynamic for different variational forms. Finally, based on our theory, I explain an implicit acceleration of using a Sobolev norm as the objective function for training.The second learning task this thesis discuss is (linear) operator learning, which has wide applications in causal inference, time series modeling, and conditional probability learning. I build the first min-max lower bound for this problem. The min-max rate has a particular structure where the more challenging parts of the input and output spaces determine the hardness of learning a linear operator. Analysis also shows that an intuitive discretization of the infinite-dimensional operator could lead to a sub-optimal statistical learning rate. Then, I'll discuss how, by suitably trading-off bias and variance, I can construct an estimator with an optimal learning rate for learning a linear operator between infinite dimension spaces. I also illustrate how this theory can inspire a multilevel machine-learning algorithm of potential practical use.
533 $a Electronic reproduction. $b Ann Arbor, Mich. : $c ProQuest, $d 2024
538 $a Mode of access: World Wide Web
650 4 $a Neural networks. $3 1011215
650 4 $a Decision making. $3 528319
655 7 $a Electronic books. $2 local $3 554714
690 $a 0800
710 2 $a Stanford University. $3 1184533
710 2 $a ProQuest Information and Learning Co. $3 1178819
773 0 $t Dissertations Abstracts International $g 85-06B.
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=30726802 $z click for full text (PQDT)