Kyutacar : Kyushu Institute of Technology Academic Repository
Not a member yet
9746 research outputs found
Sort by
Mechanism of sensory perception unveiled by simultaneous measurement of membrane voltage and intracellular calcium
Get PDFMeasuring neuronal activity is important for understanding neuronal function. Ca2+ imaging by genetically encoded calcium indicators (GECIs) is a powerful way to measure neuronal activity. Although it revealed important aspects of neuronal function, measuring the neuronal membrane voltage is important to understand neuronal function as it triggers neuronal activation. Recent progress of genetically encoded voltage indicators (GEVIs) enabled us fast and precise measurements of neuronal membrane voltage. To clarify the relation of the membrane voltage and intracellular Ca2+, we analyzed neuronal activities of olfactory neuron AWA in Caenorhabditis elegans by GCaMP6f (GECI) and paQuasAr3 (GEVI) responding to odorants. We found that the membrane voltage encodes the stimuli change by the timing and the duration by the weak semi-stable depolarization. However, the change of the intracellular Ca2+ encodes the strength of the stimuli. Furthermore, ODR-3, a G-protein alpha subunit, was shown to be important for stabilizing the membrane voltage. These results suggest that the combination of calcium and voltage imaging provides a deeper understanding of the information in neural circuits.journal articl
Self-Referential Holographic Data Storage with Integrated Denoising Function by Deep Learning
Get PDFSelf-referential holographic data storage (SR-HDS) which is one of the implementation methods of holographic data storage (HDS) enables holographic digital data recording with an one-beam optical geometry [1]．In HDS, including SR-HDS, it is desired that the datapages are reconstructed as clear as possible. For this purpose, a denoising method using deep learning for noisy reconstructed datapages where it is caused by inter-and/or intra- page interactions has recently been paid attention [2]．
In recent years, there has been a growing interest in technologies for the efficient hardware implementation of deep neural networks using optics [3]. One of them is self-referential holographic deep neural network (SR-HDNN), a method for parallel computation of deep neural networks through the application of holography [4].
Since SR-HDNN is implemented using the same optical system as SR-HDS, it is expected to implement both HDS and deep learning functions within a single optical system.
In this study, we propose to improve the quality of the reconstructed datapages of SR-HDS using the principle of SR-HDNN. Specifically, we propose the system which integrates SR-HDS and SR-HDNN and investigate its feasibility. For this purpose, we perform the numerical simulations on SR-HDS with integrated denoising function by SRHDNN.conference pape
Additional Pattern Design Method Using Deep Learning for Multi-Level Self-Referential Holographic Data Storage
Get PDFSelf-referential holographic data storage (SR-HDS) is one implementation of holographic data storage (HDS) and can record information using a single beam geometry [1]. In SR-HDS, two patterns are utilized: one is the signal pattern (SP), which is the pattern to be recorded, and the other is an arbitrary pattern, referred to as the additional pattern (AP).
For the SP to be correctly recorded, its pixel values must meet certain conditions. For example, binary patterns with the phase difference of π when phase modulated are unacceptable. Determining the pixel values for multi-level SPs can be particularly challenging. The method where SP pixel values are theoretically or numerically determined and phase-only modulation is used is referred to as the unequally spaced phase modulation (USPM) method, because the pixel values in SPs are generally unequally spaced. The complex amplitude modulation (CAM) method, on the other hand, is also acceptable. The CAM method uses binary phase modulation with m/2-level amplitude modulation to record m-value data pages [2].
On the other hand, for APs, it is beneficial to select an AP such that the spatial power spectrum of the light modulated by the sum of SP and AP becomes broader. We have proposed several methods to obtain such APs [3,4]. One method uses a searching algorithm (SA). While it has been shown that the SA-based AP design method improves reconstruction quality, it requires a significant amount of time to design an AP. Another method uses deep learning (DL), which achieves the improvement of the reconstruction quality as well as SA-based method with relatively short design time [4]. In this method, pairs of SP and AP designed by the SA-based method are used to train a deep neural network, allowing the designed AP to be quickly generated for arbitrary SPs.
One strategy to achieve high recording density in SR-HDS is to record high-quality multi-level SPs. However, the DL-based AP design method for multi-level SPs has not been explored. In this paper, we propose and numerically demonstrate the application of the DL-based AP design method to multi-level SR-HDS realized by the USPM and CAM methods.conference pape
Deep Learning-Based Resolution Enhancement of Digital Holograms Using Spatial Frequency Domain Loss Function
Get PDFSince the spatial resolution of image sensors used for digital holography is limited and is generally lower than that of photosensitive materials used for optical holography, there is a concern that the high spatial frequency component of the digital hologram is degraded. Therefore, it is desired to use an image sensor with as high a resolution as possible, however such image sensor is generally expensive and have low sensitivity due to their small pixels. To solve this problem, we have focused on the approach which appropriately interpolates the high spatial resolution component by deep learning [1].
In this study, we perform the experiment on off-axis digital holography to investigate the usefulness of deep learning for the resolution enhancement of digital holograms. Specifically, we develop a resolution-enhance deep neural network which is trained by with pairs of digital holograms acquired using image sensors with different spatial resolutions and transforms arbitrary low-resolution digital holograms into high-resolution digital holograms. In particular, we propose to use spatial frequency domain information as the loss function in the deep neural network to achieve further improvement of the resolution enhancement.conference pape
Trigger circuit design and system integration for simultaneous measurement of human EEG, motion, and gaze
Get PDFSimultaneous measurement of EEG, motion, and gaze in humans has the potential to lead to the discovery of new scientific insights. In order to achieve these simultaneous measurements, it is necessary to manage triggers and time information between measurement devices, as well as to correct time offsets. However, the management of accurate triggers and time information requires the design of a dedicated circuit board and the integration of Transistor- Transistor Logic signal voltage information. In this study, we report on the building of a trigger circuit and an experimental system using it to solve these problems. We created a home-made trigger circuit board for voltage integration and combined it with a commercially available microcomputer to realize an integrated trigger circuit and measurement system.The 2024 International Conference on Artificial Life and Robotics (ICAROB 2024), February 22-25, 2024, on line, Oita, Japanconference pape
Artificial Molecular Systems for Complex Functions Based on DNA Nanotechnology and Cell‐Sized Lipid Vesicles
Get PDFCells are highly functional and complex molecular systems. Artificially creating such systems remains a challenge, which has been extensively studied in various research fields, including synthetic biology and molecular robotics. DNA nanotechnology is a powerful tool for bottom-up engineering for constructing functional nanostructures or chemical reaction networks which can be utilized as components for artificial molecular systems. Encapsulation of these components into a giant unilamellar vesicle (GUV) composed of a lipid bilayer, the base structure of the cellular membrane, results in a functional cell-sized structure that partially mimics some cellular functions. This review discusses the studies contributing to the construction of GUV-based artificial molecular systems based on DNA nanotechnology. Molecular transport and signal transduction through lipid membranes are essential to uptake molecules from the environment and respond to stimuli. Membrane shaping relates to various functions, including motility and signaling. A chemical reaction network is required to autonomously regulate the system's functions. This review describes the functions realized using DNA nanostructures and DNA reaction networks. Given the designability and programmability of DNA nanotechnology, it may be possible that the functionality of artificial molecular systems could be comparable to or even surpass that of natural molecular systems.journal articl
The Paradox of the Coffin Trick ——Tennessee Williams's The Glass Menagerie and Hans Hofmann's Painting Theory
Get PDF本論文は「九州地区国立大学教育系・文系研究論文集」Vol.10,No.２に査読を経て受理された。journal articl
Spike-Based In-Memory Computing Circuits Using Open Source PDK
Get PDF5th International Symposium on Neuromorphic AI Hardware, March 1-2, 2024, Kyushu Institute of Technologyconference pape
Flood susceptibility mapping leveraging open-source remote-sensing data and machine learning approaches in Nam Ngum River Basin (NNRB), Lao PDR
Get PDFFrequent floods caused by monsoons and rainstorms have significantly affected the resilience of human and natural ecosystems in the Nam Ngum River Basin, Lao PDR. A cost-efficient framework integrating advanced remote sensing and machine learning techniques is proposed to address this issue by enhancing flood susceptibility understanding and informed decision-making. This study utilizes remote sensing geo-datasets and machine learning algorithms (Random Forest, Support Vector Machine, Artificial Neural Networks, and Long Short-Term Memory) to generate comprehensive flood susceptibility maps. The results highlight Random Forest’s superior performance, achieving the highest train and test Area Under the Curve of Receiver Operating Characteristic (AUROC) (1.00 and 0.993), accuracy (0.957), F1-score (0.962), and kappa value (0.914), with the lowest mean squared error (0.207) and Root Mean Squared Error (0.043). Vulnerability is particularly pronounced in low-elevation and low-slope southern downstream areas (Central part of Lao PDR). The results reveal that 36%–53% of the basin’s total area is highly susceptible to flooding, emphasizing the dire need for coordinated floodplain management strategies. This research uses freely accessible remote sensing data, addresses data scarcity in flood studies, and provides valuable insights for disaster risk management and sustainable planning in Lao PDR.journal articl
Multi-Loss Weighting for Person Re-Identification Based on Deep Learning
Get PDF視野を共有しない複数台カメラ間の人物追跡などにおいて, 同一人物を判定する人物再同定が重要である. 深層学習を用いた人物再同定は高い性能を達成しており, その損失関数として, 一般的にCross-Entropy LossやTriplet Lossが用いられる. 近年, 人物再同定に対するアプローチとして, その両方の損失関数を線形和で用いる手法が注目されている. しかし, 異なる性質を持つ損失関数を同時に用いる場合, 他方への影響を考慮した重み付き線形和による統合法が必要となる. そこで本研究では, 損失関数の学習効率の違いとそのバランスを考慮し, 学習中に損失の重みを自動調整する手法を提案する.Person re-identification is an important component to realize various image recognition systems, e.g., person tracking system by utilizing multiple cameras. Person re-identification based on deep learning has achieved high performance, and cross-entropy loss or triplet loss is generally used as the loss function. In recent years, a linear summation of both loss functions has been attracting attention as an approach to person re-identification. However, when loss functions with different properties are used at the same time, a method of synthesis by weighted linear summation that takes into account the effect of the loss function on the other loss function is necessary. To overcome above problems, in this paper, a method that automatically adjusts the weights of loss functions during learning is proposed.journal articl