Minimum Divergence Methods in Statistical Machine Learning : From an Information Geometric Viewpoint
Language: English
Published by Springer Japan, 2022
- Hardcover
- Used
Condition: Used
£ 76.49
Quantity: 1 available
Add to basketItem description from seller
Zustand: Hervorragend | Sprache: Englisch | Produktart: Bücher | This book explores minimum divergence methods of statistical machine learning for estimation, regression, prediction, and so forth, in which we engage in information geometry to elucidate their intrinsic properties of the corresponding loss functions, learning algorithms, and statistical models. One of the most elementary examples is Gauss's least squares estimator in a linear regression model, in which the estimator is given by minimization of the sum of squares between a response vector and a vector of the linear subspace hulled by explanatory vectors. This is extended to Fisher's maximum likelihood estimator (MLE) for an exponential model, in which the estimator is provided by minimization of the Kullback-Leibler (KL) divergence between a data distribution and a parametric distribution of the exponential model in an empirical analogue. Thus, we envisage a geometric interpretation of such minimization procedures such that a right triangle is kept with Pythagorean identity in the sense of the KL divergence. This understanding sublimates a dualistic interplay between a statistical estimation and model, which requires dual geodesic paths, called m-geodesic and e-geodesic paths, in a framework of information geometry. We extend such a dualistic structure of the MLE and exponential model to that of the minimum divergence estimator and the maximum entropy model, which is applied to robust statistics, maximum entropy, density estimation, principal component analysis, independent component analysis, regression analysis, manifold learning, boosting algorithm, clustering, dynamic treatment regimes, and so forth. We consider a variety of information divergence measures typically including KL divergence to express departure from one probability distribution to another. An information divergence is decomposed into the cross-entropy and the (diagonal) entropy in which the entropy associates with a generative model as a family of maximum entropy distributions; the cross entropy associates with a statistical estimation method via minimization of the empirical analogue based on given data. Thus any statistical divergence includes an intrinsic object between the generative model and the estimation method. Typically, KL divergence leads to the exponential model and the maximum likelihood estimation. It is shown that any information divergence leads to a Riemannian metric and a pair of the linear connections in the framework of information geometry. We focus on a class of information divergence generated by an increasing and convex function U, called U-divergence. It is shown that any generator function U generates the U-entropy and U-divergence, in which there is a dualistic structure between the U-divergence method and the maximum U-entropy model. We observe that a specific choice of U leads to a robust statistical procedurevia the minimum U-divergence method. If U is selected as an exponential function, then the corresponding U-entropy and U-divergence are reduced to the Boltzmann-Shanon entropy and the KL divergence; the minimum U-divergence estimator is equivalent to the MLE. For robust supervised learning to predict a class label we observe that the U-boosting algorithm performs well for contamination of mislabel examples if U is appropriately selected. We present such maximal U-entropy and minimum U-divergence methods, in particular, selecting a power function as U to provide flexible performance in statistical machine l.…
Seller Inventory # 38576074/1
- Title
- Minimum Divergence Methods in Statistical Machine Learning : From an Information Geometric Viewpoint
- Author
- Osamu Komori, Shinto Eguchi
- Publisher
- Springer Japan
- Publication year
- 2022
- Condition
- Hervorragend
- Binding
- Hardcover
- Language
- English
- ISBN 10
- 4431569200
- ISBN 13
- 9784431569206
- Seller catalogs
- Bücher
This book explores minimum divergence methods of statistical machine learning for estimation, regression, prediction, and so forth, in which we engage in information geometry to elucidate their intrinsic properties of the corresponding loss functions, learning algorithms, and statistical models. One of the most elementary examples is Gauss's least squares estimator in a linear regression model, in which the estimator is given by minimization of the sum of squares between a response vector and a vector of the linear subspace hulled by explanatory vectors. This is extended to Fisher's maximum likelihood estimator (MLE) for an exponential model, in which the estimator is provided by minimization of the Kullback-Leibler (KL) divergence between a data distribution and a parametric distribution of the exponential model in an empirical analogue. Thus, we envisage a geometric interpretation of such minimization procedures such that a right triangle is kept with Pythagorean identity in the sense of the KL divergence. This understanding sublimates a dualistic interplay between a statistical estimation and model, which requires dual geodesic paths, called m-geodesic and e-geodesic paths, in a framework of information geometry.
We extend such a dualistic structure of the MLE and exponential model to that of the minimum divergence estimator and the maximum entropy model, which is applied to robust statistics, maximum entropy, density estimation, principal component analysis, independent component analysis, regression analysis, manifold learning, boosting algorithm, clustering, dynamic treatment regimes, and so forth. We consider a variety of information divergence measures typically including KL divergence to express departure from one probability distribution to another. An information divergence is decomposed into the cross-entropy and the (diagonal) entropy in which the entropy associates with a generative model as a family of maximum entropy distributions; the cross entropy associates with a statistical estimation method via minimization of the empirical analogue based on given data. Thus any statistical divergence includes an intrinsic object between the generative model and the estimation method. Typically, KL divergence leads to the exponential model and the maximum likelihood estimation. It is shown that any information divergence leads to a Riemannian metric and a pair of the linear connections in the framework of information geometry.
We focus on a class of information divergence generated by an increasing and convex function U, called U-divergence. It is shown that any generator function U generates the U-entropy and U-divergence, in which there is a dualistic structure between the U-divergence method and the maximum U-entropy model. We observe that a specific choice of U leads to a robust statistical procedurevia the minimum U-divergence method. If U is selected as an exponential function, then the corresponding U-entropy and U-divergence are reduced to the Boltzmann-Shanon entropy and the KL divergence; the minimum U-divergence estimator is equivalent to the MLE. For robust supervised learning to predict a class label we observe that the U-boosting algorithm performs well for contamination of mislabel examples if U is appropriately selected. We present such maximal U-entropy and minimum U-divergence methods, in particular, selecting a power function as U to provide flexible performance in statistical machine learning.
"Synopsis" may belong to another edition of this title.
About the Author
Shinto Eguchi is currently Professor at the Institute of Statistical Mathematics.
"About the title" may belong to another edition of this title.
Shipping rates from Germany to U.S.A.
| Item | 60 to 60 business days | 60 to 60 business days |
|---|---|---|
| First item | £ 88.72 | £ 109.84 |
Payment methods
- Bank Wire Transfer
- Check
- Paypal
Store description
Seit über 30 Jahren handeln wir mit Büchern. Begonnen hat alles als kleiner Fachbuchhandel in Berlin. Heute sind wir spezialisiert auf den An- und Verkauf gebrauchter Bücher. Noch immer liegt unser Fokus dabei auf den Kategorien: Fachbuch und englischsprachige Literatur. Hinzu kommen Bücher aus allen anderen Genres, wie Belletristik, Biografien oder Kinderbücher. Unser oberstes Ziel ist dabei die Weiterverwendung von Dingen und die Schonung von Ressourcen. Neben dem Ankauf von privaten Büchern oder Haushaltsauflösungen kaufen wir bundesweit komplette Bibliotheken und Buchbestände von Institutionen auf, um diese nachhaltig ein weiteres Mal auf den Markt zu bringen. Parallel dazu bietet Buchpark die Möglichkeit, große Mengen einzulagern. Ganze Bibliotheken und umfangreiche Buchbestände werden von uns professionell auf Qualitätsmerkmale hin untersucht und sorgfältig für den Onlinehandel aufbereitet. Wir schenken Büchern ein zweites Leben.…
Seller's business information
Buchpark GmbH
Krügerweg 1
Trebbin, Germany 14959
Terms of sale
Buchpark
Imprint
Legal provider identification:
Buchpark GmbH
represented by the managing directors Dr. Frank Jacobi, Michael Antonius Horst Jacobi
Krügerweg 1
14959 Trebbin
Germany
Phone:+49 03381 79 76 585
E-Mail: abebooks@buchpark.de
VAT ID No.: DE243674915
registered in the commercial register of the district court Potsdam
Commercial register number HRB 33655 P
Alternative dispute resolution:
The European Commission provides a platform for out-of-court online dispute resolution (ODR platform), available at
at https://ec.europa.eu/odr.
We are a member of the "FairCommerce" initiative since 04.04.2019.
You can find more information about this at www.haendlerbund.de/faircommerce.
Cancellation policy/ sample cancellation form/
General Terms and Conditions and Customer Information/ Privacy Policy
Right of withdrawal for consumers
(A consumer is any natural person who enters into a legal transaction for purposes that are predominantly neither commercial nor self-employed).
Right of withdrawal
If you are a consumer you can withdraw from the contract in accordance with the following. Consumer means any natural person who is acting for purposes which are outside his trade, business, craft or profession.
Information regarding the right of withdrawal
Statutory right to withdraw
You have the right to withdraw from this contract within 14 days without giving any reason.
The withdrawal period will expire after 14 days from the day on which you acquire, or a third party other than the carrier and indicated by you acquires, physical possession of the last good or the last lot or piece.
To exercise the right of withdrawal, electronically fill in and submit a clear statement on our website, under "My Purchases" in "My Account". We will communicate to you an acknowledgement of receipt of such a withdrawal on a durable medium (e.g. by e-mail) without delay.
To meet the withdrawal deadline, it is sufficient for you to send your communication concerning your exercise of the right of withdrawal before the withdrawal period has expired.
Effects of withdrawal
If you withdraw from this contract, we will reimburse to you all payments received from you, including the costs of delivery (except for the supplementary costs arising if you chose a type of delivery other than the least expensive type of standard delivery offered by us).
We may make a deduction from the reimbursement for loss in value of any goods supplied, if the loss is the result of unnecessary handling by you.
We will make the reimbursement without undue delay, and not later than 14 days after the day on which we are informed about your decision to withdraw from this contract.
We will make the reimbursement using the same means of payment as you used for the initial transaction, unless you have expressly agreed otherwise; in any event, you will not incur any fees as a result of such reimbursement.
We may withhold reimbursement until we have received the goods back, or you have supplied evidence of having sent back the goods, whichever is the earliest.
You shall send back the goods or hand them over to Buchpark, Trebbin, Germany, without undue delay and in any event not later than 14 days from the day on which you communicate your withdrawal from this contract to us. The deadline is met if you send back the goods before the period of 14 days has expired. You will have to bear the direct cost of returning the goods. You are only liable for any diminished value of the goods resulting from the handling other than what is necessary to establish the nature, characteristics and functioning of the goods.
Exceptions to the right of withdrawal
The right of withdrawal does not apply to:
- The delivery of newspapers, journals or magazines with the exception of subscription contracts; and
- The supply of digital content which is not supplied on a tangible medium (e.g. on a CD or DVD) if you accepted when you placed your order that we could start to deliver it, and that you could not withdraw once delivery had started.