Committee 
Date Time 
Place 
Paper Title / Authors 
Abstract 
Paper # 
IA, ICSS 
20220624 10:25 
Nagasaki 
Univ. of Nagasaki (Primary: Onsite, Secondary: Online) 
Application of Adversarial Examples to Physical ECG Signals Taiga Ono (Waseda Univ.), Takeshi Sugawara (UEC), Jun Sakuma (Tsukuba Univ./RIKEN), Tatsuya Mori (Waseda Univ./RIKEN/NICT) IA202211 ICSS202211 
This work aims to assess the reality and feasibility of applying adversarial examples to attack cardiac diagnosis system... [more] 
IA202211 ICSS202211 pp.6166 
IBISML 
20220117 11:20 
Online 
Online 
CAMRI Loss: Classwise Additive Angular Margin Loss for Improving Recall of a Specific Class Daiki Nishiyama (Univ. Tsukuba), Fukuchi Kazuto, Yohei Akimoto, Jun Sakuma (Univ. Tsukuba/RIKEN) IBISML202122 
In realworld applications of multiclass classification models, there is a need to increase the recall of classes where ... [more] 
IBISML202122 pp.2936 
IBISML 
20220118 11:15 
Online 
Online 
[Invited Talk]
TBA Jun Sakuma (Tsukuba Univ./RIKEN) 
Explainability is one of the key elements required in medical image diagnosis using deep image recognition models. In th... [more] 

IBISML 
20220118 13:20 
Online 
Online 
IBISML202124 
We aim to explain a blackbox classifier with the form: `data X is classified as class Y because X has A, B and does not... [more] 
IBISML202124 pp.4553 
IBISML 
20181105 15:10 
Hokkaido 
Hokkaido Citizens Activites Center (Kaderu 2.7) 
[Poster Presentation]
Watermarking of Neural Network with Exponential Weighting Parameters Ryota Namba (Tsukuba Univ.), Jun Sakuma (Tsukuba Univ./riken) IBISML201863 
Deep learning has been achieving top performance in many tasks.
Since training of a deep learning model requires a gr... [more] 
IBISML201863 pp.143150 
IBISML 
20181105 15:10 
Hokkaido 
Hokkaido Citizens Activites Center (Kaderu 2.7) 
[Poster Presentation]
Differential Privacy for Likelihood Ratio Test Arashi Haishima (Univ. of Tsukuba), Jun Sakuma (Univ. of Tsukuba/RIKEN) IBISML201886 
Likelihood ratio test of logistic regression commonly used when testing whether an attribute of data significantly influ... [more] 
IBISML201886 pp.313320 
HWS, ISEC, SITE, ICSS, EMM, IPSJCSEC, IPSJSPT [detail] 
20180725 16:40 
Hokkaido 
Sapporo Convention Center 
ChiSquared Test for Independence Between Single Nucleotide Polymorphisms and Diseases Using Misoperation Resistant Searchable Homomorphic Encryption Keita Emura, Takuya Hayashi (NICT/JST), Wenjie Lu (University of Tsukuba/JST CREST), Shiho Moriai (NICT/JST), Jun Sakuma (University of Tsukuba/JST CREST/RIKEN), Yoshiji Yamada (Mie University/JST CREST) ISEC201831 SITE201823 HWS201828 ICSS201834 EMM201830 
[more] 
ISEC201831 SITE201823 HWS201828 ICSS201834 EMM201830 pp.171176 
IBISML 
20171109 13:00 
Tokyo 
Univ. of Tokyo 
Online Optimization Method for Generalized $ell_1$ Regularized Problems Yoshihiro Nakazato, Kazuto Fukuchi (Tsukuba Univ.), Jun Sakuma (Tsukuba Univ./Riken/JST) IBISML201747 
Structured sparse regularization is vital to enhance the precision and the interpretability of the model by introducing ... [more] 
IBISML201747 pp.93100 
IBISML 
20171109 13:00 
Tokyo 
Univ. of Tokyo 
Consequently Fair Contextual Bandit Learning Kazuto Fukuchi (Univ. of Tsukuba), Jun Sakuma (Univ. of Tsukuba/JST/RIKEN) IBISML201753 
Fairness in machine learning is being recognized as an important field. It requires that the consequent decisions made b... [more] 
IBISML201753 pp.139146 
IBISML 
20171109 13:00 
Tokyo 
Univ. of Tokyo 
Fast Computation of Lower Bounds for Privacy Evaluations, Based on Binary Decision Diagrams Shogo Takeuchi (Univ. of Tokyo), Kosuke Kusano, Jun Sakuma (Univ. of Tsukuba), Koji Tsuda (Univ. of Tokyo) IBISML201760 
An input value estimation is a privacy issue in a service provides information by personal information. It is necessary ... [more] 
IBISML201760 pp.193200 
IBISML 
20171110 13:00 
Tokyo 
Univ. of Tokyo 
Estimation of Training Data Distribution from Probabilistic Classifier using Generative Adversarial Networks Kosuke Kusano, Jun Sakuma (Univ. Tsukuba) IBISML201776 
Suppose we have a deep classification model $f$ that is trained with private samples that should not be released, but we... [more] 
IBISML201776 pp.301308 
IBISML 
20161116 15:00 
Kyoto 
Kyoto Univ. 
Proximal Average Accelerated Proximal Gradient Algorithm with Adaptive Restart Yoshihiro Nakazato, Kazuto Fukuchi, Jun Sakuma (Univ. Tsukuba) IBISML201655 
When using multiple regularizers, their proximal mapping is not easily available in closed form.
The method to calculat... [more] 
IBISML201655 pp.6571 
IBISML 
20161116 15:00 
Kyoto 
Kyoto Univ. 
Additive Model Decomposition with Global Sparse Structure for Multitask Granger Causal Estimation Hitoshi Abe, Jun Sakuma (Univ. Tsukuba) IBISML201656 
Causality estimation is one of the key issues in timeseries data analysis.
Granger causality is widely known as a form... [more] 
IBISML201656 pp.7379 
IBISML 
20161117 14:00 
Kyoto 
Kyoto Univ. 
Minimax optimal estimator for additively decomposable scalar functionals of discrete distributions Kazuto Fukuchi, Jun Sakuma (Univ. of Tsukuba) IBISML201682 
We deal with a problem of estimating additively decomposable scalar functionals from a set of $n$ iid samples drawn from... [more] 
IBISML201682 pp.259265 
IBISML 
20161117 14:00 
Kyoto 
Kyoto Univ. 
Empirical risk minimization for interval data and its applications to privacy preservations Hiroyuki Hanada, Toshiyuki Takada, Atsushi Shibagaki (NITech), Jun Sakuma (Univ. of Tsukuba), Ichiro Takeuchi (NITech) IBISML201689 
In this research, for machine learning tasks, we consider that the values in the training data are given as intervals an... [more] 
IBISML201689 pp.305312 
PRMU, IPSJCVIM, IBISML [detail] 
20160906 10:45 
Toyama 

A proposal on quick sensitivity analysis of empirical risk minimization problems Hiroyuki Hanada, Atsushi Shibagaki (NITech), Jun Sakuma (Univ. of Tsukuba), Ichiro Takeuchi (NITech) PRMU201680 IBISML201635 
For a training data set consisting of $n$ vectors of $d$ dimensions, we consider obtaining a training result from it by ... [more] 
PRMU201680 IBISML201635 pp.203210 
IBISML 
20151126 15:00 
Ibaraki 
Epochal Tsukuba 
[Poster Presentation]
Theoretical Vulnerability Evaluation on Linear Classifier under SelfInformation Controllable Settings Shohei Kobayashi (Univ. of Tsukuba), Shota Okumura, Ichiro Takeuchi (nitech), Jun Sakuma (Univ. of Tsukuba) IBISML201564 
Selfinformation controllable learning is machine learning algorithms that allow data providers to control their data ev... [more] 
IBISML201564 pp.8390 
IBISML 
20151127 14:00 
Ibaraki 
Epochal Tsukuba 
[Poster Presentation]
Secure Approximation Guarantee for Private Empirical Risk Minimization with Homomorphic Encryption Toshiyuki Takada, Hiroyuki Hanada (NIT), Jun Sakuma (Univ.Tsukuba), Ichiro takeuchi (NIT) IBISML201586 
Privacy concern has been increasingly important in many machine learning problems. In this paper, we study empirical ris... [more] 
IBISML201586 pp.249256 
NC, IPSJBIO, IBISML, IPSJMPS (Joint) [detail] 
20150623 13:25 
Okinawa 
Okinawa Institute of Science and Technology 
Differentially Private Multiple Hypothesis Testing Kazuya Kakizaki, Jun Sakuma (Tsukuba Univ.) IBISML20158 
Statistical hypothesis testing using test statistics ($p$value) are commonly used for identification of new scientific ... [more] 
IBISML20158 pp.4754 
NC, IPSJBIO, IBISML, IPSJMPS (Joint) [detail] 
20150623 13:50 
Okinawa 
Okinawa Institute of Science and Technology 
Lasso Granger Causality Estimation Considering Smoothness of Causality from Time Series Data Hitoshi Abe, Jun Sakuma (Tsukuba Univ.) IBISML20159 
Recently, various services for real world problems continually produce huge amount of time series data. Determination of... [more] 
IBISML20159 pp.5562 