Discriminative Perceptual Hashing (DPH) is a method that robustly verifies the identity of image content by fine-tuning a Convolutional Neural Network (CNN) for image classification to distinguish between target and non-target images. However, it does not explicitly address complex editing operations. This paper proposes a new DPH method that enables the model to verify the identity of target objects even after compositing-based modifications by adding images subjected to complex editing such as cropping the main object and pasting it into different contexts to the fine-tuning dataset. Furthermore, adjusting the types of editing operations used during training allows users to control the range of perceptual equivalence for copyright management. Experimental results demonstrate that the proposed method maintains the functionality of conventional methods and improves robustness against image compositing operations. This enhances the reliability of copyright verification under complex editing scenarios.
In this work, we consider both experimentally and numerically a bouncing-ball system comprising a ball bouncing on a table that oscillates in a sinusoidal manner. We analyze the stepwise bifurcations observed in this system through which the maximum bounce height of the ball increases in a stepwise manner as the table oscillation frequency is increased. The experiments were conducted for multiple vibration amplitudes, and we examined the changes in the ball's motion as the table frequency was varied discretely. The results show that the transitions between motion states occur in a stepwise manner with respect to the driving frequency, and that the bifurcation points obtained numerically and experimentally exhibit close quantitative agreement. Parameterizing the system in terms of the acceleration amplitude of the table, the bifurcation structure for different amplitudes collapses onto a single curve, indicating that the onset and progression of bifurcations are influenced by the acceleration amplitude rather than by the oscillation frequency alone. A phase-space analysis considering Poincare maps further reveals that each bifurcation introduces an additional state region; we find that the experimentally observed attractors appear more connected than their numerical counterparts due to the embedding reconstruction. Events associated with sticking regimes were observed in simulations but not in experiments, suggesting that small perturbations and measurement uncertainties may be sufficient to suppress such idealized behaviors.
The advent of the Modern Hopfield Network (MHN) has enabled associative memory models to handle continuous-valued data. Although MHNs exhibit high memory capacity, their recall performance tends to be unstable depending on the distribution of stored patterns. To address this issue, we propose a novel energy function based on Voronoi partitioning that enables stable memory retrieval independent of the configuration of stored patterns. Experimental results demonstrate that the proposed method achieves higher recall accuracy across a wide range of pattern sets compared with conventional MHNs.
A physical reservoir computing framework based on the Optimal Velocity Model (OVM) for ring-road traffic is proposed: OVM-RC. External inputs modulate the perceived spacing of selected vehicles, whereas others form the reservoir state. We present a theoretical spectral correspondence between the Hopf bifurcation of continuous OVM and the local stability of discrete reservoirs via an exponential map. Simulations reveal that prediction error is minimized and that traffic flow is maximized at the critical point, indicating dual optimality of computation and transport. These findings link traffic bifurcation dynamics with reservoir performance, thereby guiding the design of bifurcation-aligned physical RC systems.
The main purpose of this paper is to develop stabilization boundary control methods for the Burgers' cellular automaton and the difference Burgers' equation based on the ultradiscretization technique, and show stability analysis. We first derive a stabilizing boundary control law for the Burgers' cellular automaton, which is obtained by ultradiscretizing the Burgers' equation, and prove finite stabilization for the controlled Burgers' cellular automaton. Then, by using the inverse ultradiscretization method, we derive a stabilizing boundary control law for the difference Burgers' equation, and show stability and convergence values for the controlled difference Burgers' equation. From some numerical simulations, it turns out that the difference Burgers' equation is stabilized and the convergence value coincides with the theoretical one.
This paper describes fast image reconstruction methods for computed tomography. These methodologies employ convex programming techniques with box constraints. To accelerate these algorithms, the preconditioning matrix is introduced through the filtered back projection method. Additionally, we present the convergence theorem of the projected gradient method, which is based on a proximal gradient method, a sparse modeling technique. We show the fundamental numerical characteristics of the proposed methods and demonstrate superior image reconstruction capabilities compared to previous works in the same setting.
This paper develops a new data-driven point-to-point control method that tracks a multicopter to desired points on a given trajectory in the 3D space based on just-in-time modeling, which is one of the data-driven and model-free control methods. First, a 3D transition control method based on dual-type just-in-time modeling, which can transfer the multicopter to a desired point, is proposed. Then, by repeating the 3D transition control method, we derive a 3D point-to-point control method for the multicopter. Numerical simulations illustrate that the proposed method can realize 3D point-to-point control for the multicopter with high accuracy, and hence it has the effectiveness.
Wireless Brain-Inspired Computing (WiBIC), which implements spiking neural networks (SNNs) in a distributed manner on IoT devices, is gaining attention as a technology enabling serverless, autonomous learning. Since SNN learning relies on precise spike timing information, selecting the wireless communication protocol for transmitting spike signals is critically important. However, protocols like Carrier Sense Multiple Access with Collision Avoidance (CSMA/CA), widely used in IoT, introduce probabilistic delays to avoid collisions, potentially degrading SNN learning performance. This research proposes Asynchronous Pulse Code Multiple Access (APCMA) as a communication method suitable for spike transmission in WiBIC. In actual device experiments, CSMA/CA caused significant delay jitter and packet loss due to communication contention, whereas APCMA maintained nearly constant delay with no observed loss. For XOR learning, the learning success rate using CSMA/CA decreased as communication density increased, while APCMA maintained stable high success rate regardless of communication density.
It is known that the feature space of a Convolutional Neural Network (CNN) trained on a classification task is characterized by clusters distributed radially around the origin. In this study, to clarify the relationship between image characteristics and the norm of feature vectors, which is not directly involved in classification, we constructed a decoder to reconstruct images from these features and analyzed the feature space. Conventional decoders are trained by minimizing the Mean Squared Error (MSE) between the dataset and the reconstructed images. However, this approach does not allow for a sufficient investigation of the relationship with the norm and angle of the feature vectors. To address this, we propose a custom loss function, combining multiple loss terms, to train the decoder. Our results confirm that feature vectors with smaller norms reconstruct blurry images, whereas those with larger norms reconstruct sharp images. Furthermore, even in regions of the feature space where no training data exists, specifically on the vector of the cluster center, images were reconstructed that, despite some partial degradation, were still correctly classifiable.
In this study, we propose a bifurcation analysis method based on the monodromy matrix that can be used to analyze border-collision bifurcation points in switched dynamical systems (SDSs). First, we present the mathematical definition of an n-dimensional SDS and the behavior of solution trajectories involving switching operations. Next, we consider perturbations of solution trajectories whose initial values lie at or near to periodic points of the system, and we describe an analysis method based on the monodromy matrix that is expanded to analyze border-collision bifurcation points. Finally, a computational implementation is presented to verify the effectiveness of the proposed method.
