国際会議論文 on 池田思朗

Classification of Galactic Transients with Missing Data

Mon, 01 Sep 2025 00:00:00 +0000

Proc. ADASS (Astronomical Data Analysis Software and Systems) XXXII, pp. 348(4pp)

著者:

Yuzuki Koga
Makoto Uemura
Ryosuke Sazaki
Shiro Ikeda

URL:

Abstract:

Very early stages of galactic transients, such as nova explosions and dwarf nova outbursts, are poorly understood because their physical states change rapidly, within one day. A key problem is choosing the from different options for follow-up observations of a transient, such as imaging or spectroscopy, just after its discovery. To date, domain experts have made this choice; thus, the appropriate observations could be performed only when they were in the observatory at night. We propose an automation system that can quickly perform appropriate follow-up observations to identify the type of transients. We have developed a system called SmartK using the 1.5 m Kanata telescope of Hiroshima University. In building the system, we choose the observation type based on the mutual information of each observation mode, which is calculated with the conditional probability of the measurements obtained with the follow-up observation, and the class probability of the object. We estimate the class probability with supervised machine learning. Our training data set is characterized by many missing values, which is a common problem in the classification of astronomical objects. Sparse multinomial logistic regression (SMLR) has been used as the discriminant model in SmartK. SMLR can create nonlinear decision boundaries, although it is not easy to handle training data with missing values. Here, we define a generative model (GM) and propose a strategy to make a decision with Bayes’ theorem, and compare its performance with that of SMLR. We find that there is no significant difference in the accuracy obtained by cross-validation between SMLR and GM. It suggests that simple decision boundaries are enough for our data. We prefer GM to SMLR because it is easier to handle missing values in the training data. Additionally, GM enables us to identify anomalies that do not fall into any predefined types.

Sub/millimeter-Wave Dual-Band Line Intensity Mapping Using the Terahertz Integral Field Units with Universal Nanotechnology (TIFUUN) for the Atacama Submillimeter Telescope Experiment (ASTE)

Thu, 01 Aug 2024 00:00:00 +0000

Millimeter, Submillimeter, and Far-Infrared Detectors and Instrumentation for Astronomy XII, pp. PC1310209

著者:

Kotaro Kohno
Akira Endo
Yoichi Tamura
Akio Taniguchi
Tatsuya Takekoshi
Shiro Ikeda
Naoki Yoshida
Kana Moriwaki
Kenichi Karatsu
Jochem J. A. Baselmans
Louis H. Marting
Arend Moerman
Bruno T. Buijtendorp
Shahab Dabironezare
Matus Rybak
Tom J. L. C. Bakx
Leon G. G. Olde Scholtenhuis
Fenno Steenvoorde
Robert Huiting
David J. Thoen
Lingyu Wang
Aurora Simionescu
Stephen J. C. Yates
Alessandro Monfardini
Martino Calvo
Paul P. van der Werf
Sten Vollebregt
Bernhard R. Brandl
Tai Oshima
Ryohei Kawabe
Kazuyuki Fujita
Shunichi Nakatsubo
Yuki Kimura
Akiyoshi Tsujita
Yuki Yoshimura
Shinji Fujita
Yuri Nishimura
Yuka Yamada
Sho Fujisawa
Kanako Narita
Tetsuhiro Minamidani
Shun Ishii
Fumiya Maeda
Adam Lidz
Denis Burgarella
Bunyo Hatsukade
Fumi Egusa
Kana Morokuma-Matsui

キーワード:

integrated superconducting spectrograph (ISS)
submillimeter-wave
line-intensity mapping
ionized carbon and oxygen lines
sparse modeling
deep learning

URL:

Abstract:

We present a plan for sub/millimeter-wave line intensity mapping (LIM) using an imaging spectrograph based on the Terahertz Integral Field Units with Universal Nanotechnology (TIFUUN) architecture. We aim to measure the dust-enshrouded cosmic star formation rate density within the first 2 billion years by conducting LIM observations of ionized carbon [C II] 158 µm and oxygen [O III] 88 µm lines, redshifted to sub/millimeter wavelengths. The proposed imaging spectrograph will simultaneously observe two frequency bands: Band-1 (139–179 GHz) and Band-2 (248–301 GHz). Each band will feature up to $\sim$100 imaging pixels (spaxels), with each spaxel having 100 spectral channels, providing a modest spectral resolution (R$\sim$500). The total number of detectors (voxels) will reach $\sim$20,000. This dual-band configuration will allow simultaneous measurement of key spectral lines, e.g., [C II] 158 µm and [O III] 88 µm lines at $z = 10.2 - 12.6$, and CO(4-3), (7-6), C I and (2-1) at $z = 1.9 - 2.2$, enabling cross-correlation analysis. We will develop data-scientific methods to remove atmospheric noise using sparse modeling and to extract signals from the observed data using deep learning.

Data science based efficient and automated spectroscopy for submillimeter single-dish telescopes

Tue, 01 Aug 2023 00:00:00 +0000

2023 XXXVth General Assembly and Scientific Symposium of the International Union of Radio Science (URSI GASS), pp. 1–4

著者:

Akio Taniguchi
Yoichi Tamura
Shiro Ikeda
Tatsuya Takekoshi
Ryohei Kawabe
Kotaro Kohno
Takeshi Sakai

キーワード:

submillimeter single-dish telescope
spectroscopy

URL:

Abstract:

In this paper, we present a new data scientific approach for efficient and automated spectroscopy with millimeter and submillimeter single-dish telescopes. The proposed approach avoids direct subtraction between two noisy spectra (i.e. on-source and off-source spectra) that is common in the current data reduction: It then offers us to improve the observation sensitivity by a factor of $\gtrsim\sqrt2$ and reduce artificial baseline ripples in parallel with developing observational instruments. We demonstrate such upgrades in the real observed spectra taken by existing large millimeter single-dish telescopes. We finally discuss the application of the proposed approach for the future large submillimeter single-dish telescopes that will yield petabytes of data resulting from sensitive, wide field ($\gtrsim 1 \squaredeg$), and wide band ($>100 GHz$) imaging spectroscopy.

Image reconstruction method for an X-ray telescope with an angular resolution booster

Thu, 01 Nov 2018 00:00:00 +0000

Proc. ADASS (Astronomical Data Analysis Software and Systems) XXVIII, pp. P12.11

著者:

Mikio Morii
Shiro Ikeda
Yoshitomo Maeda

Abstract:

We propose an image reconstruction method for an X-ray telescope with an angular resolution booster proposed by Y. Maeda et al. at ISAS/JAXA. The booster consists of double coded masks in front of an X-ray mirror. In order to have a better sky image from an off-focus image, a proper image reconstruction process must be applied. The new image reconstruction method is based on the Bayesian statistics, where the traditional Richardson-Lucy algorithm is extended with a prior of sparseness and smoothness. Such a prior is desirable for astronomical imaging because astronomical objects have variety in shape from point sources, diffuse sources (supernova remnants, clusters of galaxies, and pulsar wind nebula) to mixtures of them (point sources in Galactic planes). As a result, the image resolution would be improved from a few arcmin to 10 arcsec. The performance of the X-ray telescope is demonstrated with simulated data: point sources and diffused X-ray sources such as Cas A and Crab Nebula. Through the demonstration, the angular resolution booster with the image reconstruction method is shown to be feasible.

New synthesis imaging tool for ALMA based on the sparse modeling

Thu, 01 Nov 2018 00:00:00 +0000

Proc. ADASS (Astronomical Data Analysis Software and Systems) XXVIII, pp. O4.1

著者:

Takeshi Nakazato
Shiro Ikeda
Kazunori Akiyama
George Kosugi
Masayuki Yamaguchi
Mareki Honma

Abstract:

A new imaging tool for radio interferometry has been developed based on the sparse modeling approach. It has been implemented as a Python module operating on Common Astronomy Software Applications (CASA) so that the tool is able to process the data taken by Atacama Large Millimeter/submillimeter Array (ALMA). In order to handle large data of ALMA, the Fast Fourier Transform has been implemented with gridding process. The concept of the sparse modeling for the image reconstruction has been realized with two regularization terms: L1 norm term for the sparsity and Total Squared Variation (TSV) term for the smoothness of the resulting image. Since it is important to adjust the size of the regularization terms appropriately, the cross-validation routine, which is a standard method in statistics, has been implemented. This imaging tool runs even on a standard laptop PC and processes ALMA data within a reasonable time. The interface of the tool is comprehensible to CASA users and the usage is so simple that it consists of mainly three steps to obtain the result: an initialization, a configuration, and a processing. Remarkable feature of the tool is that it produces the solution without human intervention. Furthermore, the solution is robust in the sense that it is less affected by the processing parameters. For the verification of the imaging tool, we have tested it with two extreme examples from ALMA Science Verification Data: the protoplanetary disk, HL Tau as a typical smooth and filled image, and the lensed galaxy, SDP.81 as a sparse image. In our presentation, these results will be presented with some performance information. The comparison between our results and those of traditional CLEAN method will also be provided. Finally, our future improvement and enhancement plan to make our tool competitive with CLEAN will be shown.

Qualification of Sparse Modeling Technique for radio interferometric imaging of ALMA

Thu, 01 Nov 2018 00:00:00 +0000

Proc. ADASS (Astronomical Data Analysis Software and Systems) XXVIII, pp. P12.9

著者:

George Kosugi
Takeshi Nakazato
Shiro Ikeda

Abstract:

Sparse modeling is widely used in image processing, signal processing, and machine learning recently. Thanks to the research and progress in statistical mathematics along with the evolution of computational power, the technique is supposed to be applicable to the radio interferometric imaging for the data obtained by ALMA (Atacama Large Millimeter-submillimeter Array). We’ve developed a new imaging tool based on the sparse modeling approach and experimentally implemented on CASA (Common Astronomy Software Application) which is an official reduction software for the ALMA data. The poster presentation gives supplemental information to the oral talk in session 11 by Nakazato et al. The new imaging technique with sparse modeling is a computationally intense process even for the latest CPUs. In the poster, several ideas and practices to reduce the calculation time will be presented. The comparison to CLEAN imaging by using artificial data (simulated data) will also be presented.

Super-resolution imaging of the protoplanetary disk HD 142527 using sparse modeling

Thu, 01 Nov 2018 00:00:00 +0000

Proc. ADASS (Astronomical Data Analysis Software and Systems) XXVIII, pp. P13.22

著者:

Masayuki Yamaguchi
Kazunori Akiyama
Akimasa Kataoka
Misato Fukagawa
Takashi Tsukagoshi
Takayuki Muto
Shiro Ikeda
Hiroshi Nagai
Mareki Honma
Ryohei Kawabe

Abstract:

High-resolution observations of protoplanetary disks with radio interferometers are crucial for understanding the planet formation process. Recent observations using Atacama Large Millimeter/submillimeter Array (ALMA) have revealed various small-scale structures in disks. In interferometric observations, the observed data are an incomplete set of Fourier components of the radio source image. The image reconstruction is therefore essential in obtaining the images in real space. The CLEAN technique has been widely used, but recently, a new technique using the sparse modeling approach is suggested. This technique directly solves a set of undetermined equations and has been shown to behave better than the CLEAN technique based on mock observations with VLBI (Very Long Baseline Interferometry). However, it has never been applied to ALMA-like connected interferometers nor real observational data. In this work, for the first time, the sparse modeling technique is applied to observational data sets taken by ALMA. We evaluate the performance of the technique by comparing the resulting images with those derived by the CLEAN technique. We use two sets of ALMA archival data at Band 7 ($\sim$350 GHz) for the protoplanetary disk around HD 142527. One is taken in the intermediate-baseline array configuration, and the other is in the longer-baseline array configuration. The image resolutions reconstructed from these data sets are different by a factor of $\sim 3$. We compare images reconstructed using sparse modeling and CLEAN. We find that the sparse modeling technique can successfully reconstruct the overall disk emission. The previously known disk structures appear on both images made by the sparse modeling and CLEAN at its nominal resolutions. Remarkably, the image reconstructed from intermediate-baseline data using the sparse-modeling technique matches very well with that obtained from longer-baseline data using the CLEAN technique with the accuracy of $\sim 90$% on the image domain.

New method of eclipse mapping and an application to HT Cas in the 2017 superoutburst

Sat, 01 Sep 2018 00:00:00 +0000

Proceedings of Science (The Golden Age of Cataclysmic Variables and Related Objects IV, 2017), pp. 025(7pp)

著者:

Yasuyuki Wakamatsu
Keisuke Isogai
Takashi Morita
Taichi Kato
Daisaku Nogami
Makoto Uemura
Shiro Ikeda

URL:

Abstract:

We have developed a new eclipse mapping method with Total Variation Minimization (TVM). TVM uses a concept of sparse modeling, which recovers information from sparse data. TVM sets a summation of difference in the brightness of adjacent elements in a map to be sparse. We included this concept to the eclipse mapping method and evaluated consistency of the reconstruction of the model disk. The reconstruction of the model light curve seems to be fine but that of the model disk seems to be failed, smearing the brightness distribution along the ingress/egress arcs produced by the shadow of the secondary. We applied our method to the 2017 superoutburst of HT Cas. The artifacts smearing along the ingress/egress arcs of the secondary also exist. Unaccounted noise and the short phase coverage of the input light curves seem to affect the results, leading to small artificial bright spots in the reconstructed disk.

Evaluation of large pixel CMOS image sensors for the Tomo-e Gozen wide field camera

Sun, 01 Jul 2018 00:00:00 +0000

Proc. SPIE, pp. 107091T

著者:

Yuto Kojima
Shigeyuki Sako
Ryou Ohsawa
Hidenori Takahashi
Mamoru Doi
Naoto Kobayashi
Tsutomu Aoki
Noriaki Arima
Ko Arimatsu
Makoto Ichiki
Shiro Ikeda
Kota Inooka
Yoshifusa Ita
Toshihiro Kasuga
Mitsuru Kokubo
Masahiro Konishi
Hiroyuki Maehara
Noriyuki Matsunaga
Kazuma Mitsuda
Takashi Miyata
Yuki Mori
Mikio Morii
Tomoki Morokuma
Kentaro Motohara
Yoshikazu Nakada
Shin-Ichiro Okumura
Yuki Sarugaku
Mikiya Sato
Toshikazu Shigeyama
Takao Soyano
Masaomi Tanaka
Ken’ichi Tarusawa
Nozomu Tominaga
Tomonori Totani
Seitaro Urakawa
Fumihiko Usui
Junichi Watanabe
Takuya Yamashita
Makoto Yoshikawa

URL:

Abstract:

Tomo-e Gozen (Tomo-e) is a wide field optical camera for the Kiso $1.05$ m f$/3.1$ Schmidt telescope operated by the University of Tokyo. Tomo-e is equipped with 84 chips of front-illuminated CMOS image sensors with a microlens array. The field of view is about 20 square degrees and maximum frame rate is 2 fps. The CMOS sensor has 2160$\times$1200 pixels and a size of pixel is 19 microns, which is larger than those of other CMOS sensors. We have evaluated performances of the CMOS sensors installed in Tomo-e. The readout noise is 2.0 e$^-$ in 2 fps operations when an internal amplifier gain is set to 16. The dark current is 0.5 $e^-$/sec/pix at room temperature, 290K, which is lower than a typical sky background flux in Tomo-e observations, 50 e$^-$/sec/pix. The efficiency of the camera system peaks at approximately 0.7 in 500 nm.

The Tomo-e Gozen wide field CMOS camera for the Kiso Shmidt telescope

Sun, 01 Jul 2018 00:00:00 +0000

Proc. SPIE, pp. 107020J

著者:

Shigeyuki Sako
Ryou Ohsawa
Hidenori Takahashi
Yuto Kojima
Mamoru Doi
Naoto Kobayashi
Tsutomu Aoki
Noriaki Arima
Ko Arimatsu
Makoto Ichiki
Shiro Ikeda
Kota Inooka
Yoshifusa Ita
Toshihiro Kasuga
Mitsuru Kokubo
Masahiro Konishi
Hiroyuki Maehara
Noriyuki Matsunaga
Kazuma Mitsuda
Takashi Miyata
Yuki Mori
Mikio Morii
Tomoki Morokuma
Kentaro Motohara
Yoshikazu Nakada
Shin-Ichiro Okumura
Yuki Sarugaku
Mikiya Sato
Toshikazu Shigeyama
Takao Soyano
Masaomi Tanaka
Ken’ichi Tarusawa
Nozomu Tominaga
Tomonori Totani
Seitaro Urakawa
Fumihiko Usui
Junichi Watanabe
Takuya Yamashita
Makoto Yoshikawa

URL:

Abstract:

The Tomo-e Gozen is a wide-field high-speed camera for the Kiso $1.0$ m Schmidt telescope, with a field-of-view of $20.7$ deg$^2$ covered by 84 chips of 2k$\times$1k CMOS image sensors with 19 ${\mu}$m pixels. It is capable to take consecutive images at 2 fps in full-frame read with an absolute time accuracy of 0.2 millisecond. The sensors are operated without mechanical coolers owing to a low dark current at room temperature. A low read noise of 2 e$^-$ achieves higher sensitivity than that with a CCD sensor in short exposures. Big data of 30 TBytes per night produced in the 2 fps observations is processed in real-time to quickly detect transient events and issue alerts for follow-ups.

Development of a prototype of the Tomo-e Gozen wide-field CMOS camera

Mon, 01 Aug 2016 00:00:00 +0000

Proc. SPIE, pp. 99083P

著者:

Shigeyuki Sako
Ryou Osawa
Hidenori Takahashi
Yuki Kikuchi
Mamoru Doi
Naoto Kobayashi
Tsutomu Aoki
Ko Arimatsu
Makoto Ichiki
Shiro Ikeda
Yoshifusa Ita
Toshihiro Kasuga
Hideyo Kawakita
Mitsuru Kokubo
Hiroyuki Maehara
Noriyuki Matsunaga
Hiroyuki Mito
Kazuma Mitsuda
Takashi Miyata
Kiyoshi Mori
Yuki Mori
Mikio Morii
Tomoki Morokuma
Kentaro Motohara
Yoshikazu Nakada
Kentaro Osawa
Shin-ichiro Okumura
Hiroki Onozato
Yuki Sarugaku
Mikiya Sato
Toshikazu Shigeyama
Takao Soyano
Masaomi Tanaka
Yuki Taniguchi
Ataru Tanikawa
Ken’ichi Tarusawa
Nozomu Tominaga
Tomonori Totani
Seitaro Urakawa
Fumihiko Usui
Junichi Watanabe
Jumpei Yamaguchi
Makoto Yoshikawa

URL:

Abstract:

The Tomo-e Gozen is an extremely wide-field optical camera for the Kiso 1.0-m Schmidt telescope. It is capable of taking consecutive frames with a field-of-view of 20 deg$^2$ and a sub-second time-resolution, which are achieved by 84 chips of 2k$\times$1k CMOS sensor. This camera adopts unconventional designs including a lightweight structure, a nonvacuumed and naturally-air cooled system, front-side-illuminated CMOS sensors with microlens arrays, a sensor alignment along a spherical focal plane of the telescope, and massive readout electronics. To develop technical components necessary for the Tomo-e Gozen and confirm a feasibility of its basic design, we have developed a prototype-model (PM) of the Tomo-e Gozen prior to the final-model (FM). The Tomo-e PM is equipped with eight chips of the CMOS sensor arranged in a line along the RA direction, covering a sky area of 2.0 deg$^2$. The maximum frame rate is 2 fps. The total data production rate is 80 MByte sec-1 at 2 fps, corresponding to approximately 3 TByte night$^-1$. After laboratory testing, we have successfully obtained consecutive movie data at 2 fps with the Tomo-e PM in the first commissioning run conducted in the end of 2015.

Development of a real-time data processing system for a prototype of the Tomo-e Gozen wide field CMOS camera

Mon, 01 Aug 2016 00:00:00 +0000

Proc. SPIE, pp. 991339

著者:

Ryou Ohsawa
Shigeyuki Sako
Hidenori Takahashi
Yuki Kikuchi
Mamoru Doi
Naoto Kobayashi
Tsutomu Aoki
Ko Arimatsu
Makoto Ichiki
Shiro Ikeda
Yoshifusa Ita
Toshihiro Kasuga
Hideo Kawakita
Mitsuru Kokubo
Hiroyuki Maehara
Noriyuki Matsunaga
Hiroyuki Mito
Kazuma Mitsuda
Takashi Miyata
Kiyoshi Mori
Yuki Mori
Mikio Morii
Tomoki Morokuma
Kentaro Motohara
Yoshikazu Nakada
Shin-ichiro Okumura
Hiroki Onozato
Kentaro Osawa
Yuki Sarugaku
Mikiya Sato
Toshikazu Shigeyama
Takao Soyano
Masaomi Tanaka
Yuki Taniguchi
Ataru Tanikawa
Ken’ichi Tarusawa
Nozomu Tominaga
Tomonori Totani
Seitaro Urakawa
Fumihiko Usui
Junichi Watanabe
Jumpei Yamaguchi
Makoto Yoshikawa

URL:

Abstract:

The Tomo-e Gozen camera is a next-generation, extremely wide field optical camera, equipped with 84 CMOS sensors. The camera records about a 20 square degree area at 2 Hz, providing ``astronomical movie data’’. We have developed a prototype of the Tomo-e Gozen camera (hereafter, Tomo-e PM), to evaluate the basic design of the Tomo-e Gozen camera. Tomo-e PM, equipped with 8 CMOS sensors, can capture a 2 square degree area at up to 2 Hz. Each CMOS sensor has about 2.6 M pixels. The data rate of Tomo-e PM is about 80 MB/s, corresponding to about 280 GB/hour. We have developed an operating system and reduction softwares to handle such a large amount of data. Tomo-e PM was mounted on 1.0-m Schmidt Telescope in Kiso Observatory at the University of Tokyo. Experimental observations were carried out in the winter of 2015 and the spring of 2016. The observations and software implementation were successfully completed. The data reduction is now in execution.

Sparsely Extracting Stored Movements to Construct Interfaces for Humanoid End-effector Control

Tue, 01 Dec 2015 00:00:00 +0000

Proceedings of the 2015 IEEE Conference on Robotics and Biomimetics, pp. 1816–1821

著者:

Yuka Ariki
Tetsunari Inamura
Shiro Ikeda
Jun Morimoto

URL:

Abstract:

This paper proposes a robot interface design method by which we can control humanoid end-effector movements with such a low-dimensional input device as a gamepad. In our proposed method, first, the numbers of movement trajectories to accomplish different tasks are generated using a simulated robot model and stored in a database. Second, a human user demonstrates the current task-related behavior. Third, the corresponding stored movements for the demonstrated human behavior are sparsely extracted by a sparse coding method. Finally, the sparsely extracted movement bases are linearly combined to generate a novel movement to accomplish a new target task where the linear weight parameters are modulated by the gamepad. We easily generated such complicated hand movements as spiral motions on a small humanoid robot with our proposed interface.

Sparsity and Information Processing

Tue, 01 Sep 2015 00:00:00 +0000

Symposium MEIS2015: Mathematical Progress in Expressive Image Synthesis

著者:

Shiro Ikeda

Abstract:

Recently, many information processing methods utilizing the sparsity of the information source is studied. We have reported some results on this line of research. Here we pick up two results from our own works. One is an image reconstruction method for radio interferometory and the other is a motor command computation method for a two-joint arm.

Channel capacity and achievable rates of peak power limited AWGNC, and their applications to adaptive modulation and coding

Wed, 01 Oct 2014 00:00:00 +0000

Proceedings of 2014 International Symposium on Information Theory and its Applications (ISITA2014), pp. 590–594

著者:

Shiro Ikeda
Kazunori Hayashi
Toshiyuki Tanaka

URL:

arXiv

Abstract:

The channel conditions vary over time in wireless communications. In order to transmit information efficiently, digital wireless communication systems choose the modulation scheme and coding adaptively. This framework is called the adaptive modulation and coding (AMC). The key problem of the framework is how to design the switching strategy. In this paper, we discuss the practical strategy for AMC by comparing the channel capacity, achievable rates with common modulation schemes, and the actual rates with AMC. The channel capacity is defined for a combination of the noisy channel and the constraint on the information source. The noisy channel we assume in this paper is the discrete-time complex-valued additive white Gaussian noise channel (AWGNC). For the constraint, we focus on the peak power instead of the average power since a practical communication transmitter often suffers from the peak power. We compare the capacity and achievable rates with practical modulation schemes. Furthermore, we simulate AMC and evaluate the actual rates numerically.

Compton Camera Imaging

Sun, 01 Dec 2013 00:00:00 +0000

Proceedings of 2013 Seventh International Conference on Sensing Technology (ICST2013), pp. 674–677

著者:

Shiro Ikeda
Hirokazu Odaka
Makoto Uemura
Tadayuki Takahashi
Shin Watanabe
Shin’ichiro Takeda

Abstract:

The goal of the Compton camera imaging is to visualize the gamma-ray intensity map. Here, we focus on the case where the gamma-ray sources are sufficiently far from the camera and propose a new reconstruction method for the Compton camera imaging. The method is called the bin-mode estimation (BME). The assumption is valid for astronomy applications. The method can be implemented easily, and numerical simulations show the proposed method provides sharp reconstructions.

Optimization of probability measure and its applications in information theory

Thu, 01 Aug 2013 00:00:00 +0000

Proceedings of The Sixth Workshop on Information Theoretic Methods in Science and Engineering, pp. 49

著者:

Shiro Ikeda

Abstract:

Since Shannon’s seminal work, channel capacity has been a fundamental quantity in information thoery. The general definition is formulated as an optimization problem of a probability measure under some moment constraints. Similarly, the definition of rate-distortion function is formulated as a probability measure optimization problem. We have observed that the optimal measure becomes discrete even if continuous measures are allowed. The same phenomena are observed in Bayesian statistics, where the channel capacity problem is deeply related to the reference prior. In this talk, the background of the problem is introduced with some examples, and its applications to communication theory is discussed.

Rate-distortion function for gamma sources under absolute-log distortion measure

Mon, 01 Jul 2013 00:00:00 +0000

Proceedings of 2013 IEEE International Symposium on Information Theory (ISIT2013), pp. 2557–2561

著者:

Kazuho Watanabe
Shiro Ikeda

URL:

Abstract:

We evaluate the rate-distortion function for the i.i.d. gamma sources with respect to the absolute-log distortion measure. The logarithmic transformation reduces this rate-distortion problem to that under the absolute distortion measure. Extending the explicit evaluation of the rate-distortion function for the Gaussian sources, we obtain the parametric form of the rate-distortion function. We show that the optimal distribution of reconstruction consists of a continuous component enclosed by left and right discrete components and the left discrete component vanishes when the allowed distortion is small.

Convex formulation for nonparametric estimation of mixing distribution

Wed, 01 Aug 2012 00:00:00 +0000

Proceedings of The Fifth Workshop on Information Theoretic Methods in Science and Engineering, pp. 36–39

著者:

Kazuho Watanabe
Shiro Ikeda

Abstract:

We discuss a nonparametric estimation method of the mixing distribution in mixture models. We propose an objective function with one parameter, where its minimization becomes the maximum likelihood estimation or the kernel vector quantization in special cases. Generalizing Lindsay’s theorem for the nonparametric maximum likelihood estimation, we prove the existence and discreteness of the optimal mixing distribution and devise an algorithm to calculate it. Furthermore, we show the connection between the unifying estimation framework and the rate-distortion problem. It is demonstrated that with an appropriate choice of the parameter, the proposed method is less prone to overfitting than the maximum likelihood method.

Sparse Phase Retrieval

Wed, 01 Jun 2011 00:00:00 +0000

Proceedings of 4th Workshop on Signal Processing with Adaptive Sparse Structured Representations (SPARS2011), pp. 106

著者:

Shiro Ikeda
Hidetoshi Kono

Abstract:

Coherent X-ray Diffraction Imaging (CXDI) is a technique for the 2-dimensional (2D) and 3D reconstruction of nanoscale structures. The detector receives the photons scattered by the object, and ideally, the diffraction pattern gives the power spectrum of the electron density. Since we are only provided the power spectrum and the phase is lost, we need to retrieve the phase in order to reconstruct the structure from the diffraction image.

Motor Planning as an Optimization of Command Representation

Tue, 01 Dec 2009 00:00:00 +0000

Proceedings of 48th IEEE Conference on Decision and Control (CDC2009), pp. 4499–4504

著者:

Shiro Ikeda
Yutaka Sakaguchi

URL:

Abstract:

A fundamental problem in the field of motor neuroscience is to understand how our brain generates appropriate motor commands for precise movements effortlessly. The problem seems difficult since there are infinitely many possible trajectories and our musculo-skeltal system is generally redundant. We focus on the motor command representation and show that a simple strategy can solve the problem for a planar two-joints arm model. We also discuss the emergence of the muscle synergies, which may enable us to make natural motor behaviors with smaller degrees of freedom.

Spiking Neuron Channel

Thu, 01 Jan 2009 00:00:00 +0000

Proceedings of 2009 IEEE International Symposium on Information Theory (ISIT2009), pp. 1589–1593

著者:

Shiro Ikeda
Jonathan H. Manton

キーワード:

channel capacity
channel capacity achieving distribution
neuroscience
neuron

URL:

Abstract:

The information transfer through a single neuron is a fundamental information processing in the brain. This paper studies the information-theoretic capacity of a single neuron by treating the neuron as a communication channel. Two different models are considered. The temporal coding model of a neuron as a communication channel assumes the output is $\tau$ where $\tau$ is a gamma-distributed random variable corresponding to the inter-spike interval, that is, the time it takes for the neuron to fire once. The rate coding model is similar; the output is the actual rate of firing over a fixed period of time. We prove that for both models, the capacity achieving distribution has only a finite number of probability mass points. This allows us to compute numerically the capacity of a neuron. Our capacity results are in a plausible range based on biological evidence to date.

Capacity of a single neuron channel

Sat, 01 Nov 2008 00:00:00 +0000

3rd Australian Workshop on Mathematical and Computational Neuroscience, NeuroEng 2008, pp. 13

著者:

Shiro Ikeda
Jonathan H. Manton

Abstract:

The information transfer through a single neuron is a fundamental information processing in the brain and computing the information channel capacity is important to understand the information processing in the brain. In this work, we discuss the capacity of a single spiking neuron channel. The problem is difficult since the capacity depends on various issues, such as coding, characteristics of the communication channel and optimisation over input distributions. In this work, two different coding models are considered. The temporal coding model of a neuron as a communication channel assumes the output is $\tau$ where $\tau$ is a gamma-distributed random variable corresponding to the inter-spike interval, that is, the time it takes for the neuron to fire once. The rate coding model is similar; the output is the actual rate of firing over a fixed period of time. Theoretical studies prove that the distribution of inputs, which achieves the channel capacity, is a discrete distribution with finite mass points for temporal and rate coding under a reasonable assumption. This allows us to compute numerically the capacity of a neuron. Numerical results are in a plausible range based on biological evidence to date.

A game theoretical analysis of combining classifiers for multi-class classification problems

Mon, 01 Oct 2007 00:00:00 +0000

Proceedings of the international workshop on data-mining and statistical science (DMSS2007), pp. 149–160

著者:

Yuichi Shiraishi
Kenji Fukumizu
Shiro Ikeda

URL:

Abstract:

Combining binary classifiers for multi-class classification problems has been very popular after the invention of SVM and ada-boost, which are known to be very effective for binary classification. In this paper, we analyze theoretically the ECOC approach, which is a standard combining method. We discuss the problem of combining binary classifiers from the game-theoretical point of view. First, we develop a general theorem for the condition of minimaxity, which is closely related to the network flow theory. Applying this theorem, we show that the ECOC approach has the minimax property in the one-vs-one and one-vs-all case.

Improving mobile reception of digital satellite broadcasting

Sat, 01 Sep 2007 00:00:00 +0000

Proceeding of The 18th Annual IEEE International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC'07), pp. 1–5

著者:

Masatoshi Hamada
Shiro Ikeda

URL:

Abstract:

This article proposes a new method to improve the reception of the digital satellite broadcasting on a moving vehicle. There are two major factors which make the reception difficult, the Doppler shift and the fading channel. The proposed method does not rely on any diversity receptions nor pilot sequences but uses the channel estimation and stochastic inference. First, Doppler shift is estimated and removed. Then the fading channel is represented with a graphical model, and the parameters are estimated by maximum likelihood estimation. The inference of the code words based on the graphical model is computed and used for decoding. We confirmed the improvements through experiments with numerical simulations and real data. Computational cost is reasonable for real time system and the results are promising.

Motor Planning Based on Sparse Command Representation

Fri, 01 Sep 2006 00:00:00 +0000

Brain-inspired Information Technology (BrainIT2006), pp. P1-8

著者:

Yutaka Sakaguchi
Shiro Ikeda

Abstract:

Recent computational studies have succeeded in presenting trajectory planning algorithms which could well replicate human movements. However, most of them have not explicitly clarified the problem of motor command representation. The authors recently proposed a novel motor planning algorithm based on ``sparse representation,’’ which tries to design motor commands with combination of a smaller number of basis patterns. This paper presents generalized formulation of this framework.

Motor Planning and Sparse Motor Command Representation

Sat, 01 Jul 2006 00:00:00 +0000

Processingds of Fifteenth Annual Computational Neuroscience Meeting (CNS$\ast$2006), pp. 89

著者:

Yutaka Sakaguchi
Shiro Ikeda

Abstract:

The present article proposes a novel computational approach to the motor planning. In the proposed approach, each motor command is represented as a linear combination of prefixed basis patterns, and the command for a given task is designed by minimizing a two-termed ``information representation criterion’’ which consists of a task optimization term and a parameter preference term. The result of a computer simulation with a single-joint reaching task confirmed that the proposed framework appropriately worked, together with showing that the resultant trajectory qualitatively replicated Fitts’ law.

Sparse representation and piece-wise linear kernel

Tue, 01 Nov 2005 00:00:00 +0000

Proceedings of Signal Processing with Adaptive Sparse Structured Representations (SPARS'05), pp. 123–126

著者:

Shiro Ikeda

Abstract:

We propose a new type of kernel function, where feature space is explicitly given with a piece-wise linear mapping from the input space. This idea is inspired by sparse linear system analysis, where inputs are represented as a sparse linear combination of ``dictionary vectors.’’ This article gives the idea of such kernel function, and some preliminary experimental results.

Information geometry of turbo and LDPC codes

Tue, 01 Mar 2005 00:00:00 +0000

Proceedings of 2005 IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP2005)

著者:

Shiro Ikeda

URL:

Abstract:

Turbo and LDPC (Low-Density Parity Check) codes are simple and new type of error correction codes which give a powerful and practical performance of error correction. Although experimental results show their efficacy, further theoretical analysis is necessary, which is not straightforward. We have built unified framework of turbo and LDPC codes based on information geometry. The framework helps our intuitive understanding of the codes and opens a new prospect of further analysis. We have revealed some properties of these codes in the proposed framework. This paper summarizes the results.

Information Geometry of Loopy BP

Sun, 01 Jun 2003 00:00:00 +0000

Supplementary Proceedings of ICANN/ICONIP2003, pp. 54–57

著者:

Shiro Ikeda
Toshiyuki Tanaka
Shun-ichi Amari

Abstract:

Belief propagation (BP) is an efficient algorithm to solve the inference problem of graphical models. We give the information geometrical view of the algorithm, and propose a new cost function which yields a new algorithm.

Information Geometry of Turbo Codes

Sat, 01 Jun 2002 00:00:00 +0000

Proceedings of 2002 IEEE International Symposium on Information Theory (ISIT2002), pp. 114–119

著者:

Shiro Ikeda
Toshiyuki Tanaka
Shun-ichi Amari

URL:

Abstract:

The properties of the turbo decoding is studied from information geometrical viewpoint. Our study gives an intuitive understanding of the theoretical background, and a new framework for the analysis. Based on the framework, we reveal basic properties of the turbo decoding.

Information Geometrical Framework for Analyzing Belief Propagation Decoder

Mon, 01 Apr 2002 00:00:00 +0000

Advances in Neural Information Processing Systems, pp. 407–414

著者:

Shiro Ikeda
Toshiyuki Tanaka
Shun-ichi Amari

URL:

Link

Abstract:

The mystery of belief propagation (BP) decoder, especially of the turbo decoding, is studied from information geometrical viewpoint. The loopy belief network (BN) of turbo codes makes it difficult to obtain the true ``belief’’ by BP, and the characteristics of the algorithm and its equilibrium are not clearly understood. Our study gives an intuitive understanding of the mechanism, and a new framework for the analysis. Based on the framework, we reveal basic properties of the turbo decoding.

Information-geometrical significance of sparsity in Gallager codes

Mon, 01 Apr 2002 00:00:00 +0000

Advances in Neural Information Processing Systems, pp. 527–534

著者:

Toshiyuki Tanaka
Shiro Ikeda
Shun-ichi Amari

URL:

Link

Abstract:

We report a result of perturbation analysis on decoding error of the belief propagation decoder for Gallager code. The analysis is based on information geometry, and it shows that the principal term of decoding error at equilibrium comes from the m-embedding curvature of the log-linear submanifold spanned by the estimated pseudoposteriors, one for the full marginal, and K for the posteriors with single checks, where K is the number of checks in the Gallager code. It is then shown that the principal error term vanishes when the parity-check matrix of the code is so sparse that there are no two columns with overlap greater than 1.

Belief propagation and turbo code: Information geometrical view

Thu, 01 Nov 2001 00:00:00 +0000

Proceedings of 2001 International Conference on Neural Information Processing (ICONIP2001), pp. 41–46

著者:

Shiro Ikeda
Toshiyuki Tanaka
Shun-ichi Amari

Abstract:

In this article, we describe the information geometrical understanding of the belief propagation decoder, especially of the turbo decoding. The turbo decoding was proposed by Berrou et al. early in 90’s, and many studies have been appeared on this practical and powerful error correcting code. Even though many experimental results support the potential of the turbo decoding, there is not sufficient theoretical analysis for the decoding method. We investigate the problem from information geometrical viewpoint. From the new viewpoint, we establish a new framework for analyzing the turbo code, and reveal basic properties.

Blind source separation in reflective sound fields

Tue, 01 May 2001 00:00:00 +0000

Proceedings of 2001 IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP2001)

著者:

Futoshi Asano
Shiro Ikeda
Michiaki Ogawa
Hideki Asoh
Nobuhiko Kitawaki

URL:

Abstract:

Two array signal processing techniques are combined with independent component analysis to enhance the performance of blind separation of acoustic signals in a reflective environment such as rooms. The first technique is the subspace method which reduces the effect of room reflection. The second technique is a method of solving the permutation, in which the coherency of the mixing matrix in adjacent frequencies is utilized.

Blind source separation in reflective sound fields

Sun, 01 Apr 2001 00:00:00 +0000

Proceedings of Workshop on Hands-Free Speech Communication 2001 (HSC2001), pp. 51–54

著者:

Futoshi Asano
Shiro Ikeda
Michiaki Ogawa
Hideki Asoh
Nobuhiko Kitawaki

Abstract:

In this paper, the effect of room reflaction on blind source separation is investigated. The higher order reflection (reverberation) can be reduced in advance of blind separation by using the subspace method. On the other hand, as for the lower order reflection (early reflection), it is shown by experiments that the early reflection has little effect on the separation performance.

Motion Adaptation Depresses Perception of Contour from Motion and Magnetoencephalography (MEG) Responses in V2/3 and V5

Wed, 01 Nov 2000 00:00:00 +0000

30th Annual meeting. Society for Nueroscience Abstracts (SfN2000), pp. 670

著者:

Masashi Tanigawa
Kenji Yoshikawa
Keisuke Toyama
Yoshio Ohtani
Yoshimichi Ejima
Shiro Ikeda
Shigeki Kajihara

Abstract:

Signals of contextual modulation (CM) underlying perception of motion from contour (CFM) have been studied as difference between cortical responses to segmented motion (SM) where random dots (RDs) in the neighboring segments moved in the opposite direction and homogenous motion (HM) where all RDs moved in the same direction (Lamme et al., 1994). Using a magnetometer-spatial filter system (Shimadzu SBI-100) capable of resolving individual visual cortical activities, we found similar CM signals localized in human V2/3. The CM signals occurred behind V5 responses, suggesting that CFM is conducted in V2/3, instructed by global motion signals detected in V5 and back-conveyed to V2/3 (Toyama et al., 1999). The present study reports two findings supporting this view. First, a fine parallelism was found between CM signals in V2/3 and CFM determined by the forced choice estimate for mixed SM (MSM) where SM was modified to contain variable fractions (7-50%) of oppositely moving RDs. Visual stimuli were presented to the right visual field of 7 male subjects, and MEGs were recorded from the left occipital cortex. In all subjects, CFM started to decrease with 10% MSM and reduced to 35% with 40% MSM. There was a corresponding reduction (35%) in CM signals in V2/3. This parallelism was even preserved after adaptation to HM (for 3min.). CFM decreased dramatically after adaptation (65% reduction for 40% MSM), paralleled by a similar robust reduction (35%) of the CM signals. Second, there was also a dramatic reduction of motion perception (80% for 70% MHM) for mixed HM (MHM) where HM was modified to contain various fractions (70 – 90%) of incoherent RD motion, and a reduction of V5 response (40%), but practically no reduction in V1 and V2/3 responses ($<$10%). These findings support the above view and further indicate that V5 is the primary site for motion perception and adaptation.

Separation of MEG Signal by Independent Component Analysis

Tue, 01 Aug 2000 00:00:00 +0000

Proceedings of the 12th International Conference on Biomagnetism (Biomag2000)

著者:

Shigeki Kajihara
Keisuke Toyama
Shiro Ikeda
Noboru Murata
Tatsuya Kobayashi
Yoshihisa Kida
Yoshikazu Yoshida

Abstract:

Independent component analysis (ICA) is a way to resolve signals into independent components based on the statistical characteristics of the signals. Magnetoencephalograms (MEGs) containing the brain signals as well as the noise from various sources such as the electrocardiograms (ECG), electromyograms, the earth magnetism and the device noises which are mutually independent may be as appropriate target for ICA. We applied ICA to the epileptic and the visual evoked fields (VEF) for separation of noise and signal components.

ICA for noisy neurobiological data

Sat, 01 Jul 2000 00:00:00 +0000

Proceedings of International Joint Conference on Neural Networks 2000 (IJCNN2000)

著者:

Shiro Ikeda
Keisuke Toyama

URL:

Abstract:

ICA (Independent Component Analysis) is a new technique for analyzing multi-variant data. Lots of results are reported in the field of neurobiological data analysis such as EEG (Electroencephalography), MRI (Magnetic Resonance Imaging), and MEG (Magnetoencephalography) using ICA. But there still remain problems. In most of the neurobiological data, there are a large amount of noise, and the number of independent components is unknown which gives difficulties for many ICA algorithms. In this article, we discuss an approach to separate noise-contaminated data without knowing the number of independent components. The idea is to replace PCA (Principal Component Analysis), which is used as the preprocessing of many ICA algorithms, with factor analysis. In the new preprocessing, the number of the sources and the amount of the noise are estimated. After the preprocessing, an ICA algorithm is used to estimate the separation matrix and mixing system. Through the experiments with MEG data, we show this approach is effective.

Evaluation and Real-time implementation of blind source separation system using time-delayed decorrelation

Thu, 01 Jun 2000 00:00:00 +0000

Proceedings of International Workshop on Independent Component Analysis and Blind Signal Separation (ICA2000), pp. 411–415

著者:

Futoshi Asano
Shiro Ikeda

Abstract:

Blind source separation based on the time-delayed decorrelation, which had been extended to a convolved mixture problem, is evaluated and implemented in a DSP system. The evalutation is conducted using the data recorded in an anechoic chamber and those in a listening room with moderate reverberation. The crosstalks and the scores of automatic speech recognition are measured. The hardware consists of dual DSPs and a host PC. The results of the benchmark is shown.

Factor Analysis Preprocessing for ICA

Thu, 01 Jun 2000 00:00:00 +0000

Proceedings of International Workshop on Independent Component Analysis and Blind Signal Separation (ICA2000), pp. 327–332

著者:

Shiro Ikeda

Abstract:

One of the reasons ICA (Independent Component Analysis) became so popular is that ICA is a promising tools for a lot of applications. One of the attractive applications is the biological data analysis. There are a lot of works on neurobiological data analysis such as EEG (Electroencephalography), fMRI (functional Magnetic Resonance Imaging), and MEG (Magnetoencephalography), and they show interesting results. However, there still remain some problems to be solved. Many neurobiological data includes a large amount of noises, and also the number of independent components is unknown. These problems make it difficult to obtain good results by ICA algorithms. We discuss an approach to separate the data which contain additive noise without knowing the number of independent components. Our approach uses factor analysis as the preprocessing of the ICA algorithm, instead of PCA (Principal Component Analysis), which is the major preprocessing in many ICA algorithms. In the new preprocessing, the number of the sources and the amount of sensor noise are estimated. After the preprocessing, an ICA algorithm is used to estimate the separation matrix and mixing system. Through the experiments with MEG data and fMRI data, we show this approach is effective.

Application of independent component analysis (ICA) to magnetoencephalography

Fri, 01 Oct 1999 00:00:00 +0000

29th Annual meeting. Society for Nueroscience Abstracts (SfN1999), pp. 1930

著者:

Shigeki Kajihara
Kenji Yoshikawa
Noboru Murata
Shiro Ikeda
Masashi Tanigawa
Yoshio Ohtani
Yoshimichi Ejima
Keisuke Toyama

キーワード:

neuroscience
ICA
MEG

Abstract:

Independent component analysis (ICA) is a method to reconstruct individual source signals from the complex signal containing signals and noises from various sources, based on the second or higher order statistics of the complex signal, and has been found to be useful for improvement of signal to noise (S/N) ratio in electroencephalograms (EEGs) and other electrophysiological recordings. We applied ICA to magnetoencephalograms (MEGs) for separation of MEG signals arising from different cortical areas as well as for improvement of S/N ratio. MEG signals recorded from 4 male subjects using a 129 channel magnetometer focused to the left occipital cortex during presentation of visual stimuli (stationary sinusoidal grating and moving random dots) to the right visual field. A demixing matrix was determined so that 129 ICA components conveyed the minimum cross-correlation (a range of time delay, 0–70 ms). Less than 10 of the entire components were usually synchronized to the visual stimuli, and the original MEGs were reconstructed by inverse-transforming those components through the inverse demixing matrix. There was a dramatic improvement in the S/N ratio (10 times). In addition, analysis by spatial filters transforming MEGs into local currents in the brain (Toyama et al., 1999) revealed that it was possible to identify responses arising from different cortical areas based on the temporal patterns of the ICA components. Stationary sinusoidal grating produced a single response localized in VI, while moving random-dots produced multiple responses with different time courses localized in VI, V2/3 and V5, respectively. Supported by Grant-Aid from the Japanese Ministry of Edu. Sci. and Cul. and Special Coordination Funds from the Sci. and Tech. Agency of the Japanese government, Grant in Aid for Scientific Research on Priority Area, Research for the future 1996L00201.

Combining Independent Component Analysis and Sound Stream Segregation

Sun, 01 Aug 1999 00:00:00 +0000

Proceedings of IJCAI-99 Workshop on Computational Auditory Scene Analysis, pp. 92–98

著者:

Hiroshi G. Okuno
Shiro Ikeda
Tomohiro Nakatani

キーワード:

sound separation
ICA
signal processing

Abstract:

This paper reports the issues and results of AI Challenge: ``Understanding Three Simultaneous Speeches’’. First, the issues of the Challenge are revisited. We emphasis the importance of information fusion of various attributes of speeches (sounds) in separating speeches from a mixture of sounds. This emphasis is supported by comparing two methods of speech separation; computational auditory scene analysis approach that employs the attributes of sound sources and sound transmitting channel, and blind source separation approach that dispenses with these attributes. Although these two approaches are usually considered as opposite with regards to whether sound attributes is used or not, we conclude that they differ in the ways of using sound attributes. Next, a new algorithm for information fusion is proposed. Sound attributes extracted by tracking harmonic structures and sound source directions as well as by independent component analysis are fused according to sound ontology. Finally, the error reduction rate of the 1-best/10-best word recognition of each speaker performed on 200 mixtures of two women’s and one man’s utterances of an isolated word is reported.

Convergence of the Wake-Sleep Algorithm

Tue, 01 Jun 1999 00:00:00 +0000

Advances in Neural Information Processing Systems, pp. 239–245

著者:

Shiro Ikeda
Shun-ichi Amari
Hiroyuki Nakahara

キーワード:

EM algorithm
wake-sleep algorithm
information geometry

URL:

Link

Abstract:

The W-S (Wake-Sleep) algorithm is a simple learning rule for the models with hidden variables. It is shown that this algorithm can be applied to a factor analysis model which is a linear version of the Helmholtz machine. But even for a factor analysis model, the general convergence is not proved theoretically. In this article, we describe the geometrical understanding of the W-S algorithm in contrast with the EM (Expectation-Maximization) algorithm and the em algorithm. As the result, we prove the convergence of the W-S algorithm for the factor analysis model. We also show the condition for the convergence in general models.

A method of ICA in time-frequency domain

Fri, 01 Jan 1999 00:00:00 +0000

Proceedings of International Workshop on Independent Component Analysis and Blind Signal Separation (ICA'99), pp. 365–371

著者:

Shiro Ikeda
Noboru Murata

キーワード:

sound separation
ICA
signal processing

Abstract:

We propose a method of ICA for separating convolutive mixtures of acoustic signals. The acoustic signals recorded in a real environment are not instantaneous but convolutive mixtures, because of the delay and the reflections. In order to separate these signals, it is effective to transform the signals into time-frequency domain. The difficult point in these approaches is the ambiguity of the permutation and amplitude which is unavoidable in original ICA problem. Since the basic ICA approaches cannot solve these ambiguity, we need another approach to solve them. We employed the envelopes of the signals to solve it, and have developed some algorithms. In this article, we show the outline of our original method, and some extensions of it. They are, the on-line version and auditory scene analysis problem.

A method of blind separation based on temporal structure of signals

Thu, 01 Oct 1998 00:00:00 +0000

Proceedings of 1998 International Conference on Neural Information Processing (ICONIP'98), pp. 737–742

著者:

Shiro Ikeda
Noboru Murata

キーワード:

sound separation
ICA
signal processing

Abstract:

In this article, we propose an Blind Source Separation algorithm for convolutive mixture of signals. We propose a method of separating signals in the time-frequency domain. We apply the decorrelation method proposed by Molgedey and Schuster on spectrogram and reconstruct separated signals focusing on the temporal structure of the signals. We show some results of experiments with both artificially controlled data and speech data recorded in the real environment.

A On-line Algorithm for Blind Source Separation on Speech Signals

Tue, 01 Sep 1998 00:00:00 +0000

Proceedings of 1998 International Symposium on Nonlinear Theory and its Applications (NOLTA'98), pp. 923–926

著者:

Noboru Murata
Shiro Ikeda

キーワード:

sound separation
ICA
signal processing

Abstract:

In this article, we propose an on-line algorithm for Blind Source Separation of speech signals, which is recorded in a real environment. This on-line algorithm makes it possible to trace the changing environment. The idea is to apply some on-line algorithm in the time-frequency domain. We show some results of experiments.

An Approach to Blind Source Separation of Speech Signals

Tue, 01 Sep 1998 00:00:00 +0000

Proceedings of 1998 International Conference on Artificial Neural Networks (ICANN'98), pp. 761–766

著者:

Shiro Ikeda
Noboru Murata

キーワード:

sound separation
ICA
signal processing

Abstract:

In this paper we introduce a new technique for blind source separation of speech signals. We focused on the temporal structure of signals which is not always the case in other major approaches. The idea is to apply the decorrelation method proposed by Molgedey and Schuster in time-frequency domain. We show some results of experiments with artificial data and speech data recorded in the real environment. Our algorithm needs considerably straightforward calculation and includes only a few parameters to be tuned.

Acceleration of the EM algorithm

Mon, 01 Dec 1997 00:00:00 +0000

Proceedings of 1997 International Symposium on Nonlinear Theory and its Applications (NOLTA'97), pp. 755–758

著者:

Shiro Ikeda

キーワード:

EM algorithm

Abstract:

The EM algorithm is widely used to estimate the parameters of many applications. It is simple but the convergence speed is slow. There is another algorithm called the scoring method which is faster but complicated. We show these two methods can be connected by using the EM algorithm recursively.

A Self-Organizing System with Cell-Specialization

Tue, 01 Apr 1997 00:00:00 +0000

Proceedings of 1997 IEEE International Conference on Evolutionary Computing (ICEC'97), pp. 279–284

著者:

Kaoru Nakano
Katsumi Konishi
Rui Ishiyama
Shiro Ikeda

キーワード:

evolutionary computing

URL:

Abstract:

For the study of evolutionary systems, we have already proposed several models. In this paper, we are concerned with a different type of evolutionary system, the cell-specialization. As seen in hydras, cell-specialization is one of the very important types of self-organization. In the process of cell-specialization, each cell which has the same DNA is specialized. The whole system which consists of these cells has a highly advanced function. We propose a kind of coding (something like DNA), ``a system description’’ which makes this possible. Using the system description, we built a hardware model which shows the validity for self-organization, and also did some computer simulations which shows this can be used for evolutionary systems.

Construct The Structure of Stochastic Multilayer Perceptron Using The Model Search Method

Tue, 01 Oct 1996 00:00:00 +0000

Proceedings of 1996 International Symposium on Nonlinear Theory and its Applications (NOLTA'96), pp. 197–200

著者:

Shiro Ikeda

キーワード:

model selection

Abstract:

The author gives an algorithm to search the structure of a stochastic models with hidden variable. The author have shown the algorithm to find the hidden structure of the Hidden Markov Model and in this article, the algorithm is applied for one of the other stochastic models which have hidden probabilistic variables.

A Learning Machine That Evolves

Fri, 01 Dec 1995 00:00:00 +0000

Proceedings of 1995 IEEE International Conference on Evolutionary Computing (ICEC'95), pp. 808–813

著者:

Kaoru Nakano
Hideaki Hiraki
Shiro Ikeda

キーワード:

Perceptron
evolutionary computing

URL:

Abstract:

We propose a simple model of a learning machine that evolves. When a classification problem is given, a Perceptron like learning machine obtains a proper set of feature detecting cells through mating, mutation, and natural selection. Computer simulation showed the expected results. This is one of our trials to approach the evolutionary system in the real world.

Estimate The Source Structure Through Communication

Fri, 01 Dec 1995 00:00:00 +0000

Proceedings of 1995 IEEE International Conference on Neural Networks (ICNN'95), pp. 2799–2802

著者:

Shiro Ikeda
Kaoru Nakano

キーワード:

model selection

URL:

Abstract:

In order to categorize a set of data which consists of some categories, it is important to know the probability distribution of the data of each category. Using these probability distributions, one can classify the data. In most cases, such as speech and image recognition problems, the data, for training are categorized in advance. But there are some cases when only the uncategorized data are available. For example, when a baby learns phonation, it is hard to give it the samples of each phone separately, it learns the number of the phones and how to phonate each of them through observing only the uncategorized data and making communication with the parents or the teacher. In this article, the authors give an algorithm for this situation. In the algorithm, the distribution of whole data is described with a finite mixture model. The model can observe only the uncategorized data but can make a kind of communication with the teacher (which is the probability source). The parameters of the model are estimated using the EM algorithm and the number of the categories are determined through a communication with the teacher. A numerical simulation of a simple image recognition problem is given.

Construction of Phoneme Models –- Model Search of Hidden Markov Models –-

Fri, 01 Oct 1993 00:00:00 +0000

Proceedings of International Workshop on Intelligent Signal Processing and Communication Systems (ISPACS'93), pp. 82–87

著者:

Shiro Ikeda

キーワード:

hidden Markov model
model selection

Abstract:

The author proposes an algorithm to define a structure of a HMM (hidden Markov model). HMMs are widely used in the speech recognition systems and at that time structures are usually determined according to the heuristic knowledge. In this article this problem is treated as so-called ``model selection’’ problem in statistics. Two recognition experiments using this algorithm are shown. First, artificial data then, ATR speech database are used for the source. Through these experiments, the author shows that such model selection is effective.

A Robot Organizing Purposive Behavior by Itself

Mon, 01 Jun 1992 00:00:00 +0000

Proceedings of International Joint Conference on Neural Networks (IJCNN'92), pp. 570–575

著者:

Shiro Ikeda
Kaoru Nakano
Yutaka Sakaguchi

キーワード:

self-organization
associative memory
neural networks

URL:

Abstract:

For studying the mechanism of the brain, so called ``synthetic approach’’ is effective. Synthetic approach is to conjecture the mechanism of the target through constructing its model. We have constructed some twenty models of the brain for this study. In this article, we describe one of them which we constructed recently. The model includes abilities of perception, memory, and action. To have these three abilities enables the model to realize highly intellectual behavior or self-organizing ability that cannot be realized by a model having just one ability. We realized the model in the form of a robot which organizes purposive behavior by itself. This robot forms effective behavioral patterns to achieve the purpose through trial and error.

国際会議論文 on 池田 思朗

Classification of Galactic Transients with Missing Data

著者:

URL:

Abstract:

Sub/millimeter-Wave Dual-Band Line Intensity Mapping Using the Terahertz Integral Field Units with Universal Nanotechnology (TIFUUN) for the Atacama Submillimeter Telescope Experiment (ASTE)

著者:

キーワード:

URL:

Abstract:

Data science based efficient and automated spectroscopy for submillimeter single-dish telescopes

著者:

キーワード:

URL:

Abstract:

Image reconstruction method for an X-ray telescope with an angular resolution booster

著者:

Abstract:

New synthesis imaging tool for ALMA based on the sparse modeling

著者:

Abstract:

Qualification of Sparse Modeling Technique for radio interferometric imaging of ALMA

著者:

Abstract:

Super-resolution imaging of the protoplanetary disk HD 142527 using sparse modeling

著者:

Abstract:

New method of eclipse mapping and an application to HT Cas in the 2017 superoutburst

著者:

URL:

Abstract:

Evaluation of large pixel CMOS image sensors for the Tomo-e Gozen wide field camera

著者:

URL:

Abstract:

The Tomo-e Gozen wide field CMOS camera for the Kiso Shmidt telescope

著者:

URL:

Abstract:

Development of a prototype of the Tomo-e Gozen wide-field CMOS camera

著者:

URL:

Abstract:

Development of a real-time data processing system for a prototype of the Tomo-e Gozen wide field CMOS camera

著者:

URL:

Abstract:

Sparsely Extracting Stored Movements to Construct Interfaces for Humanoid End-effector Control

著者:

URL:

Abstract:

Sparsity and Information Processing

著者:

Abstract:

Channel capacity and achievable rates of peak power limited AWGNC, and their applications to adaptive modulation and coding

著者:

URL:

Abstract:

Compton Camera Imaging

著者:

Abstract:

Optimization of probability measure and its applications in information theory

著者:

Abstract:

Rate-distortion function for gamma sources under absolute-log distortion measure

著者:

URL:

Abstract:

Convex formulation for nonparametric estimation of mixing distribution

著者:

Abstract:

Sparse Phase Retrieval

著者:

Abstract:

Motor Planning as an Optimization of Command Representation

著者:

URL:

Abstract:

Spiking Neuron Channel

著者:

国際会議論文 on 池田思朗