# Kei Takeuchi

## Profiles.Takeuchi History

May 15, 2017, at 09:01 AM by Vovk - Akimichi's corrections
Changed line 6 from:
to:
Changed line 10 from:
* 1994 - 2008  Professor, Faculty of International Studies, Meiji Gakuin University
to:
* 1994 - 2006  Professor, Faculty of International Studies, Meiji Gakuin University
Changed lines 23-25 from:
In 1979 Kei Takeuchi wrote a handout (dated July 17 1979) for a seminar at Stanford university in which he anticipated some features of conformal prediction.  Developing the notion of an $1-\alpha$ expectation tolerance region, he posed the problem of finding the smallest (with respect to the Lebesgue measure) such region that would be valid under the [[Main/randomness assumption]].  Invoking the standard completeness result for the order statistics, he replaced the [[Main/randomness assumption]] by the [[Main/exchangeable probability distribution | exchangeability assumption]].  Assuming that the observations are generated from a probability distribution with density $f_0$, he showed that the optimal solution is the [[Main/conformal predictor]] assigning the conformity score $f_0(z_i)$ to an observation $z_i$.  This is a special case of [[Main/conformal prediction]] not involving learning, but he also developed a more general version that does require some learning (namely, parameter estimation).  If the true distribution is only supposed to belong to a parametric statistical model, he suggested using the plug-in density for computing the conformity scores $f_0(z_i\mid\hat\theta)$.  To preserve validity, $\hat\theta$ was to be computed from the extended training sequence with the test object assigned a postulated label $y$, making Prof. Takeuchi's procedure an instance of [[Main/conformal prediction]].

[[Attach:790717-takeuchi-memo.pdf]]
to:
In 1979 Kei Takeuchi wrote a handout (dated July 17 1979, [[(Attach:)790717-takeuchi-memo.pdf]]) for a seminar at Stanford university in which he anticipated some features of conformal prediction.  Developing the notion of an $1-\alpha$ expectation tolerance region, he posed the problem of finding the smallest (with respect to the Lebesgue measure) such region that would be valid under the [[Main/randomness assumption]].  Invoking the standard completeness result for the order statistics, he replaced the [[Main/randomness assumption]] by the [[Main/exchangeable probability distribution | exchangeability assumption]].  Assuming that the observations are generated from a probability distribution with density $f_0$, he showed that the optimal solution is the [[Main/conformal predictor]] assigning the conformity score $f_0(z_i)$ to an observation $z_i$.  This is a special case of [[Main/conformal prediction]] not involving learning, but he also developed a more general version that does require some learning (namely, parameter estimation).  If the true distribution is only supposed to belong to a parametric statistical model, he suggested using the plug-in density for computing the conformity scores $f_0(z_i\mid\hat\theta)$.  To preserve validity, $\hat\theta$ was to be computed from the extended training sequence with the test object assigned a postulated label $y$, making Prof. Takeuchi's procedure an instance of [[Main/conformal prediction]].
May 14, 2017, at 04:31 PM by Vovk - attached Takeuchi's notes
Changed lines 23-25 from:
In 1979 Kei Takeuchi wrote a handout (dated July 17 1979) for a seminar at Stanford university in which he anticipated some features of conformal prediction.  Developing the notion of an $1-\alpha$ expectation tolerance region, he posed the problem of finding the smallest (with respect to the Lebesgue measure) such region that would be valid under the [[Main/randomness assumption]].  Invoking the standard completeness result for the order statistics, he replaced the [[Main/randomness assumption]] by the [[Main/exchangeable probability distribution | exchangeability assumption]].  Assuming that the observations are generated from a probability distribution with density $f_0$, he showed that the optimal solution is the [[Main/conformal predictor]] assigning the conformity score $f_0(z_i)$ to an observation $z_i$.  This is a special case of [[Main/conformal prediction]] not involving learning, but he also developed a more general version that does require some learning (namely, parameter estimation).  If the true distribution is only supposed to belong to a parametric statistical model, he suggested using the plug-in density for computing the conformity scores $f_0(z_i\mid\hat\theta)$.  To preserve validity, $\hat\theta$ was to be computed from the extended training sequence with the test object assigned a postulated label $y$, making Prof. Takeuchi's procedure an instance of [[Main/conformal prediction]].
to:
In 1979 Kei Takeuchi wrote a handout (dated July 17 1979) for a seminar at Stanford university in which he anticipated some features of conformal prediction.  Developing the notion of an $1-\alpha$ expectation tolerance region, he posed the problem of finding the smallest (with respect to the Lebesgue measure) such region that would be valid under the [[Main/randomness assumption]].  Invoking the standard completeness result for the order statistics, he replaced the [[Main/randomness assumption]] by the [[Main/exchangeable probability distribution | exchangeability assumption]].  Assuming that the observations are generated from a probability distribution with density $f_0$, he showed that the optimal solution is the [[Main/conformal predictor]] assigning the conformity score $f_0(z_i)$ to an observation $z_i$.  This is a special case of [[Main/conformal prediction]] not involving learning, but he also developed a more general version that does require some learning (namely, parameter estimation).  If the true distribution is only supposed to belong to a parametric statistical model, he suggested using the plug-in density for computing the conformity scores $f_0(z_i\mid\hat\theta)$.  To preserve validity, $\hat\theta$ was to be computed from the extended training sequence with the test object assigned a postulated label $y$, making Prof. Takeuchi's procedure an instance of [[Main/conformal prediction]].

[[Attach:790717-takeuchi-memo.pdf]]
Changed line 23 from:
In 1979 Kei Takeuchi wrote a handout (dated July 17 1979) in which he anticipated some features of conformal prediction.  Developing the notion of an $1-\alpha$ expectation tolerance region, he posed the problem of finding the smallest (with respect to the Lebesgue measure) such region that would be valid under the [[Main/randomness assumption]].  Invoking the standard completeness result for the order statistics, he replaced the [[Main/randomness assumption]] by the [[Main/exchangeable probability distribution | exchangeability assumption]].  Assuming that the observations are generated from a probability distribution with density $f_0$, he showed that the optimal solution is the [[Main/conformal predictor]] assigning the conformity score $f_0(z_i)$ to an observation $z_i$.  This is a special case of [[Main/conformal prediction]] not involving learning, but he also developed a more general version that does require some learning (namely, parameter estimation).  If the true distribution is only supposed to belong to a parametric statistical model, he suggested using the plug-in density for computing the conformity scores $f_0(z_i\mid\hat\theta)$.  To preserve validity, $\hat\theta$ was to be computed from the extended training sequence with the test object assigned a postulated label $y$, making Prof. Takeuchi's procedure an instance of [[Main/conformal prediction]].
to:
In 1979 Kei Takeuchi wrote a handout (dated July 17 1979) for a seminar at Stanford university in which he anticipated some features of conformal prediction.  Developing the notion of an $1-\alpha$ expectation tolerance region, he posed the problem of finding the smallest (with respect to the Lebesgue measure) such region that would be valid under the [[Main/randomness assumption]].  Invoking the standard completeness result for the order statistics, he replaced the [[Main/randomness assumption]] by the [[Main/exchangeable probability distribution | exchangeability assumption]].  Assuming that the observations are generated from a probability distribution with density $f_0$, he showed that the optimal solution is the [[Main/conformal predictor]] assigning the conformity score $f_0(z_i)$ to an observation $z_i$.  This is a special case of [[Main/conformal prediction]] not involving learning, but he also developed a more general version that does require some learning (namely, parameter estimation).  If the true distribution is only supposed to belong to a parametric statistical model, he suggested using the plug-in density for computing the conformity scores $f_0(z_i\mid\hat\theta)$.  To preserve validity, $\hat\theta$ was to be computed from the extended training sequence with the test object assigned a postulated label $y$, making Prof. Takeuchi's procedure an instance of [[Main/conformal prediction]].
Changed line 23 from:
In 1977 Kei Takeuchi wrote a handout (dated July 17 1979) in which he anticipated some features of conformal prediction.  Developing the notion of an $1-\alpha$ expectation tolerance region, he posed the problem of finding the smallest (with respect to the Lebesgue measure) such region that would be valid under the [[Main/randomness assumption]].  Invoking the standard completeness result for the order statistics, he replaced the [[Main/randomness assumption]] by the [[Main/exchangeable probability distribution | exchangeability assumption]].  Assuming that the observations are generated from a probability distribution with density $f_0$, he showed that the optimal solution is the [[Main/conformal predictor]] assigning the conformity score $f_0(z_i)$ to an observation $z_i$.  This is a special case of [[Main/conformal prediction]] not involving learning, but he also developed a more general version that does require some learning (namely, parameter estimation).  If the true distribution is only supposed to belong to a parametric statistical model, he suggested using the plug-in density for computing the conformity scores $f_0(z_i\mid\hat\theta)$.  To preserve validity, $\hat\theta$ was to be computed from the extended training sequence with the test object assigned a postulated label $y$, making Prof. Takeuchi's procedure an instance of [[Main/conformal prediction]].
to:
In 1979 Kei Takeuchi wrote a handout (dated July 17 1979) in which he anticipated some features of conformal prediction.  Developing the notion of an $1-\alpha$ expectation tolerance region, he posed the problem of finding the smallest (with respect to the Lebesgue measure) such region that would be valid under the [[Main/randomness assumption]].  Invoking the standard completeness result for the order statistics, he replaced the [[Main/randomness assumption]] by the [[Main/exchangeable probability distribution | exchangeability assumption]].  Assuming that the observations are generated from a probability distribution with density $f_0$, he showed that the optimal solution is the [[Main/conformal predictor]] assigning the conformity score $f_0(z_i)$ to an observation $z_i$.  This is a special case of [[Main/conformal prediction]] not involving learning, but he also developed a more general version that does require some learning (namely, parameter estimation).  If the true distribution is only supposed to belong to a parametric statistical model, he suggested using the plug-in density for computing the conformity scores $f_0(z_i\mid\hat\theta)$.  To preserve validity, $\hat\theta$ was to be computed from the extended training sequence with the test object assigned a postulated label $y$, making Prof. Takeuchi's procedure an instance of [[Main/conformal prediction]].
Changed line 23 from:
In 1977 Kei Takeuchi wrote a handout (dated July 17 1979) in which he anticipated some features of conformal prediction.  Developing the notion of an $1-\alpha$ expectation tolerance region, he posed the problem of finding the smallest (with respect to the Lebesgue measure) such region that would be valid under the [[Main/randomness assumption]].  Invoking the standard completeness result for the order statistics, he replaced the [[Main/randomness assumption]] by the [[Main/exchangeable probability distribution | exchangeability assumption]].  Assuming that the observations are generated from a probability distribution with density $f_0$, he showed that the optimal solution is the [[Main/conformal predictor]] assigning the conformity score $f_0(z_i)$ to an observation $z_i$.  This is a special case of [[Main/conformal prediction]] not involving learning, but he also developed a more general version that does require some learning (namely, parameter estimation).  If the true distribution is only supposed to belong to a parametric statistical model, he suggested using the plug-in density for computing the conformity scores $f_0(z_i\mid\hat\theta)$.  To preserve validity $\theta$ was to be computed from the extended training sequence with the test object assigned a postulated label $y$, making Prof. Takeuchi's procedure an instance of conformal prediction.
to:
In 1977 Kei Takeuchi wrote a handout (dated July 17 1979) in which he anticipated some features of conformal prediction.  Developing the notion of an $1-\alpha$ expectation tolerance region, he posed the problem of finding the smallest (with respect to the Lebesgue measure) such region that would be valid under the [[Main/randomness assumption]].  Invoking the standard completeness result for the order statistics, he replaced the [[Main/randomness assumption]] by the [[Main/exchangeable probability distribution | exchangeability assumption]].  Assuming that the observations are generated from a probability distribution with density $f_0$, he showed that the optimal solution is the [[Main/conformal predictor]] assigning the conformity score $f_0(z_i)$ to an observation $z_i$.  This is a special case of [[Main/conformal prediction]] not involving learning, but he also developed a more general version that does require some learning (namely, parameter estimation).  If the true distribution is only supposed to belong to a parametric statistical model, he suggested using the plug-in density for computing the conformity scores $f_0(z_i\mid\hat\theta)$.  To preserve validity, $\hat\theta$ was to be computed from the extended training sequence with the test object assigned a postulated label $y$, making Prof. Takeuchi's procedure an instance of [[Main/conformal prediction]].
Changed line 23 from:
In 1977 Kei Takeuchi wrote a handout (dated July 17 1979) in which he anticipated some features of conformal prediction.  Developing the notion of an $1-\alpha$ expectation tolerance region, he posed the problem of finding the smallest (with respect to the Lebesgue measure) such region that would be valid under the [[Main/randomness assumption]].  Invoking the standard completeness result for the order statistics, he replaced the [[Main/randomness assumption]] by the [[Main/exchangeable probability distribution | exchangeability assumption]].  Assuming that the observations are generated from a probability distribution with density $f_0$, he showed that the optimal solution is the [[Main/conformal predictor]] assigning the conformity score $f_0(z_i)$ to an observation $z_i$.  This is a special case of [[Main/conformal prediction]] not involving learning, but he developed a more general version that does require some learning (namely, parameter estimation).  If the true distribution is only supposed to belong to a parametric statistical model, he suggested using the plug-in density for computing the conformity scores $f_0(z_i\mid\hat\theta)$.  To preserve validity $\theta$ was to be computed from the extended training sequence with the test object assigned a postulated label $y$, making Prof. Takeuchi's procedure an instance of conformal prediction.
to:
In 1977 Kei Takeuchi wrote a handout (dated July 17 1979) in which he anticipated some features of conformal prediction.  Developing the notion of an $1-\alpha$ expectation tolerance region, he posed the problem of finding the smallest (with respect to the Lebesgue measure) such region that would be valid under the [[Main/randomness assumption]].  Invoking the standard completeness result for the order statistics, he replaced the [[Main/randomness assumption]] by the [[Main/exchangeable probability distribution | exchangeability assumption]].  Assuming that the observations are generated from a probability distribution with density $f_0$, he showed that the optimal solution is the [[Main/conformal predictor]] assigning the conformity score $f_0(z_i)$ to an observation $z_i$.  This is a special case of [[Main/conformal prediction]] not involving learning, but he also developed a more general version that does require some learning (namely, parameter estimation).  If the true distribution is only supposed to belong to a parametric statistical model, he suggested using the plug-in density for computing the conformity scores $f_0(z_i\mid\hat\theta)$.  To preserve validity $\theta$ was to be computed from the extended training sequence with the test object assigned a postulated label $y$, making Prof. Takeuchi's procedure an instance of conformal prediction.
Changed line 23 from:
In 1977 Kei Takeuchi wrote a handout (dated July 17 1979) in which he anticipated some features of conformal prediction.  Developing the notion of an $1-\alpha$ expectation tolerance region, he posed the problem of finding the smallest (with respect to the Lebesgue measure) such region that would be valid under the [[randomness assumption]].  Invoking the standard completeness result for the order statistics, he replaced the [[randomness assumption]] by the [[exchangeable probability distribution | exchangeability assumption]].  Assuming that the observations are generated from a probability distribution with density $f_0$, he showed that the optimal solution is the [[conformal predictor]] assigning the conformity score $f_0(z_i)$ to an observation $z_i$.  This is a special case of [[conformal prediction]] not involving learning, but he developed a more general version that does require some learning (namely, parameter estimation).  If the true distribution is only supposed to belong to a parametric statistical model, he suggested using the plug-in density for computing the conformity scores $f_0(z_i\mid\hat\theta)$.  To preserve validity $\theta$ was to be computed from the extended training sequence with the test object assigned a postulated label $y$, making Prof. Takeuchi's procedure an instance of conformal prediction.
to:
In 1977 Kei Takeuchi wrote a handout (dated July 17 1979) in which he anticipated some features of conformal prediction.  Developing the notion of an $1-\alpha$ expectation tolerance region, he posed the problem of finding the smallest (with respect to the Lebesgue measure) such region that would be valid under the [[Main/randomness assumption]].  Invoking the standard completeness result for the order statistics, he replaced the [[Main/randomness assumption]] by the [[Main/exchangeable probability distribution | exchangeability assumption]].  Assuming that the observations are generated from a probability distribution with density $f_0$, he showed that the optimal solution is the [[Main/conformal predictor]] assigning the conformity score $f_0(z_i)$ to an observation $z_i$.  This is a special case of [[Main/conformal prediction]] not involving learning, but he developed a more general version that does require some learning (namely, parameter estimation).  If the true distribution is only supposed to belong to a parametric statistical model, he suggested using the plug-in density for computing the conformity scores $f_0(z_i\mid\hat\theta)$.  To preserve validity $\theta$ was to be computed from the extended training sequence with the test object assigned a postulated label $y$, making Prof. Takeuchi's procedure an instance of conformal prediction.
May 14, 2017, at 10:44 AM by Vovk - added information about his 1977 notes
Changed lines 21-23 from:
* [[Main/Game-theoretic Probability]]
to:
* [[Main/Game-theoretic Probability]]

In 1977 Kei Takeuchi wrote a handout (dated July 17 1979) in which he anticipated some features of conformal prediction.  Developing the notion of an $1-\alpha$ expectation tolerance region, he posed the problem of finding the smallest (with respect to the Lebesgue measure) such region that would be valid under the [[randomness assumption]].  Invoking the standard completeness result for the order statistics, he replaced the [[randomness assumption]] by the [[exchangeable probability distribution | exchangeability assumption]].  Assuming that the observations are generated from a probability distribution with density $f_0$, he showed that the optimal solution is the [[conformal predictor]] assigning the conformity score $f_0(z_i)$ to an observation $z_i$.  This is a special case of [[conformal prediction]] not involving learning, but he developed a more general version that does require some learning (namely, parameter estimation).  If the true distribution is only supposed to belong to a parametric statistical model, he suggested using the plug-in density for computing the conformity scores $f_0(z_i\mid\hat\theta)$.  To preserve validity $\theta$ was to be computed from the extended training sequence with the test object assigned a postulated label $y$, making Prof. Takeuchi's procedure an instance of conformal prediction.
April 20, 2008, at 01:46 PM by Volodya Vovk -
Changed line 1 from:
(:title Professor Kei Takeuchi:)
to:
(:title Kei Takeuchi:)
April 20, 2008, at 01:43 PM by Volodya Vovk -
Changed lines 1-2 from:
Kei Takeuchi holds PhD in Economics.  Until recently he was Professor at the Faculty of International Studies, MeijiGakuin University.  He is also Professor Emeritus, University of Tokyo;
Director, Institute for International Studies, MeijiGakuin University; Chairman, Statistics Council of Japan; President, Japan Statistical Association; President, Center for Academic Publications Japan; Former President, Japan Statistical Society.
to:
(:title Professor Kei Takeuchi:)
Kei Takeuchi holds PhD in Economics.  Until recently he was Professor at the Faculty of International Studies, MeijiGakuin University.  He is also Professor Emeritus, University of
Tokyo; Director, Institute for International Studies, MeijiGakuin University; Chairman, Statistics Council of Japan; President, Japan Statistical Association; President, Center for Academic Publications Japan; Former President, Japan Statistical Society.
April 16, 2008, at 10:25 AM by Volodya Vovk -
Changed lines 5-13 from:
* 1933      Born in Tokyo, Japan
* 1963-1994  Associate Professor then Professor, Faculty of Economics, University of Tokyo
* 1987-1994  Professor, Research Center for Advanced Science and Technology (Joint Appointment)
* 1994
Retired from the University of Tokyo
* 1994-2008  Professor, Faculty
of International Studies, Meiji Gakuin University
* 1988-1994  Member
of Science Council of Japan
* 1997-2000
Member of Science Council of Japan
to:
* 1933        Born in Tokyo, Japan
* 1963 - 1994  Associate Professor then Professor, Faculty of Economics, University of Tokyo
* 1987 - 1994  Professor, Research Center for Advanced Science and Technology (Joint Appointment)
* 1994
Retired from the University of Tokyo
* 1994 - 2008  Professor, Faculty
of International Studies, Meiji Gakuin University
* 1988 - 1994  Member of Science Council of Japan
* 1997 -
2000  Member of Science Council of Japan
Changed lines 17-18 from:
* Japanese Economy,
* Socio-Economic Impacts of Science and technology
to:
* Japanese Economy
* Socio-Economic Impacts of Science and Technology
April 16, 2008, at 10:23 AM by Volodya Vovk - created profile based on an OCDE web page
Kei Takeuchi holds PhD in Economics.  Until recently he was Professor at the Faculty of International Studies, MeijiGakuin University.  He is also Professor Emeritus, University of Tokyo;
Director, Institute for International Studies, MeijiGakuin University; Chairman, Statistics Council of Japan; President, Japan Statistical Association; President, Center for Academic Publications Japan; Former President, Japan Statistical Society.

Personal History:
* 1933      Born in Tokyo, Japan
* 1963-1994  Associate Professor then Professor, Faculty of Economics, University of Tokyo
* 1987-1994  Professor, Research Center for Advanced Science and Technology (Joint Appointment)
* 1994      Retired from the University of Tokyo
* 1994-2008  Professor, Faculty of International Studies, Meiji Gakuin University
* 1988-1994  Member of Science Council of Japan
* 1997-2000  Member of Science Council of Japan

Research Areas:
* Mathematical Statistics
* Econometrics
* Japanese Economy,
* Socio-Economic Impacts of Science and technology
* General History of Civilization
* Global Environment Problems
* [[Main/Game-theoretic Probability]]