Phenetic Research Articles

Learning nonparametric ordinary differential equations from noisy data

Learning nonparametric systems of Ordinary Differential Equations (ODEs) x˙=f(t,x) from noisy data is an emerging machine learning topic. We use the well-developed theory of Reproducing Kernel Hilbert Spaces (RKHS) to define candidates for f for which the solution of the ODE exists and is unique. Learning f consists of solving a constrained optimization problem in an RKHS. We propose a penalty method that iteratively uses the Representer theorem and Euler approximations to provide a numerical solution. We prove a generalization bound for the L2 distance between x and its estimator. Experiments are provided for the FitzHugh–Nagumo oscillator, the Lorenz system, and for predicting the Amyloid level in the cortex of aging subjects. In all cases, we show competitive results compared with the state-of-the-art.

Down-sampling in diffusion MRI: a bundle-specific DTI and NODDI study.

Multi-shell diffusion Magnetic Resonance Imaging (dMRI) data has been widely used to characterise white matter microstructure in several neurodegenerative diseases. The lack of standardised dMRI protocols often implies the acquisition of redundant measurements, resulting in prolonged acquisition times. In this study, we investigate the impact of the number of gradient directions on Diffusion Tensor Imaging (DTI) and on Neurite Orientation Dispersion and Density Imaging (NODDI) metrics. Data from 124 healthy controls collected in three different longitudinal studies were included. Using an in-house algorithm, we reduced the number of gradient directions in each data shell. We estimated DTI and NODDI measures on six white matter bundles clinically relevant for neurodegenerative diseases. Fractional Anisotropy (FA) measures on bundles where data were sampled at the 30% rate, showed a median L1 distance of up to 3.92% and a 95% CI of (1.74, 8.97)% when compared to those obtained at reference sampling. Mean Diffusivity (MD) reached up to 4.31% and a 95% CI of (1.60, 16.98)% on the same premises. At a sampling rate of 50%, we obtained a median of 3.90% and a 95% CI of (1.99, 16.65)% in FA, and 5.49% with a 95% CI of (2.14, 21.68)% in MD. The Intra-Cellular volume fraction (ICvf) median L1 distance was up to 2.83% with a 95% CI of (1.98, 4.82)% at a 30% sampling rate and 3.95% with a 95% CI of (2.39, 7.81)% at a 50% sampling rate. The volume difference of the reconstructed white matter at reference and 50% sampling reached a maximum of (2.09 ± 0.81)%. In conclusion, DTI and NODDI measures reported at reference sampling were comparable to those obtained when the number of dMRI volumes was reduced by up to 30%. Close to reference DTI and NODDI metrics were estimated with a significant reduction in acquisition time using three shells, respectively with: 4 directions at a b value of 700 s/mm2, 14 at 1000 s/mm2, and 32 at 2000 s/mm2. The study revealed aspects that can be important for large-scale clinical studies on bundle-specific diffusion MRI.

Multi-dimensional multi-round minimum cost consensus models with iterative mechanisms involving reward and punishment measures

In order to address the issues of unfairness in consensus and subjectivity bias in data, while further expanding the traditional minimum cost consensus model (MCCM) in terms of opinion dimension and consensus round, we propose the multi-dimensional MCCM (MD-MCCM) and multi-round MCCM (MR-MCCM). Subsequently, we introduce the concepts and definitions pertaining to these two variants of consensus and formulate their respective consensus model frameworks. To benefit from the advantages of both, we integrate these two types of consensus models, thereby introducing the concept of multi-dimensional multi-round minimum cost consensus. To address the issues of unfairness and subjectivity in the consensus-reaching process, we propose two specific reward and punishment measures and devise the multi-dimensional multi-round consensus iterative mechanisms. These mechanisms facilitate an iterative update of decision-makers’ opinions, weights, and unit costs within the feedback process. Based on this foundation, we construct the multi-dimensional multi-round MCCM (MDMR-MCCM) involving Euclidean and Manhattan distance measures, and their corresponding algorithm and consensus reaching process are presented. Finally, a group decision problem concerning site selection validates the rationality and effectiveness of our models. Moreover, a comparative analysis of results obtained using different models is provided.

Foreseeing Reconstruction Quality of Gradient Inversion: An Optimization Perspective

Gradient inversion attacks can leak data privacy when clients share weight updates with the server in federated learning (FL). Existing studies mainly use L2 or cosine distance as the loss function for gradient matching in the attack. Our empirical investigation shows that the vulnerability ranking varies with the loss function used. Gradient norm, which is commonly used as a vulnerability proxy for gradient inversion attack, cannot explain this as it remains constant regardless of the loss function for gradient matching. In this paper, we propose a loss-aware vulnerability proxy (LAVP) for the first time. LAVP refers to either the maximum or minimum eigenvalue of the Hessian with respect to gradient matching loss at ground truth. This suggestion is based on our theoretical findings regarding the local optimization of the gradient inversion in proximity to the ground truth, which corresponds to the worst case attack scenario. We demonstrate the effectiveness of LAVP on various architectures and datasets, showing its consistent superiority over the gradient norm in capturing sample vulnerabilities. The performance of each proxy is measured in terms of Spearman's rank correlation with respect to several similarity scores. This work will contribute to enhancing FL security against any potential loss functions beyond L2 or cosine distance in the future.

Progressively Knowledge Distillation via Re-parameterizing Diffusion Reverse Process

Knowledge distillation aims at transferring knowledge from the teacher model to the student one by aligning their distributions. Feature-level distillation often uses L2 distance or its variants as the loss function, based on the assumption that outputs follow normal distributions. This poses a significant challenge when distribution gaps are substantial since this loss function ignores the variance term. To address the problem, we propose to decompose the transfer objective into small parts and optimize it progressively. This process is inspired by diffusion models from which the noise distribution is mapped to the target distribution step by step. However, directly employing diffusion models is impractical in the distillation scenario due to its heavy reverse process. To overcome this challenge, we adopt the structural re-parameterization technique to generate multiple student features to approximate the teacher features sequentially. The multiple student features are combined linearly in inference time without extra cost. We present extensive experiments performed on various transfer scenarios, such as CNN-to-CNN and Transformer-to-CNN, that validate the effectiveness of our approach.

Transformer-based multimodal change detection with multitask consistency constraints

Change detection plays a fundamental role in Earth observation for analyzing temporal iterations over time. However, recent studies have largely neglected the utilization of multimodal data that presents significant practical and technical advantages compared to single-modal approaches. This research focuses on leveraging pre-event digital surface model (DSM) data and post-event digital aerial images captured at different times for detecting change beyond 2D. We observe that the current change detection methods struggle with the multitask conflicts between semantic and height change detection tasks. To address this challenge, we propose an efficient Transformer-based network that learns shared representation between cross-dimensional inputs through cross-attention. It adopts a consistency constraint to establish the multimodal relationship. Initially, pseudo-changes are derived by employing height change thresholding. Subsequently, the L2 distance between semantic and pseudo-changes within their overlapping regions is minimized. This explicitly endows the height change detection (regression task) and semantic change detection (classification task) with representation consistency. A DSM-to-image multimodal dataset encompassing three cities in the Netherlands was constructed. It lays a new foundation for beyond-2D change detection from cross-dimensional inputs. Compared to five state-of-the-art change detection methods, our model demonstrates consistent multitask superiority in terms of semantic and height change detection. Furthermore, the consistency strategy can be seamlessly adapted to the other methods, yielding promising improvements.

LDMVFI: Video Frame Interpolation with Latent Diffusion Models

Existing works on video frame interpolation (VFI) mostly employ deep neural networks that are trained by minimizing the L1, L2, or deep feature space distance (e.g. VGG loss) between their outputs and ground-truth frames. However, recent works have shown that these metrics are poor indicators of perceptual VFI quality. Towards developing perceptually-oriented VFI methods, in this work we propose latent diffusion model-based VFI, LDMVFI. This approaches the VFI problem from a generative perspective by formulating it as a conditional generation problem. As the first effort to address VFI using latent diffusion models, we rigorously benchmark our method on common test sets used in the existing VFI literature. Our quantitative experiments and user study indicate that LDMVFI is able to interpolate video content with favorable perceptual quality compared to the state of the art, even in the high-resolution regime. Our code is available at https://github.com/danier97/LDMVFI.

Project-Fair and Truthful Mechanisms for Budget Aggregation

We study the budget aggregation problem in which a set of strategic voters must split a finite divisible resource (such as money or time) among a set of competing projects. Our goal is twofold: We seek truthful mechanisms that provide fairness guarantees to the projects. For the first objective, we focus on the class of moving phantom mechanisms, which are -- to this day -- essentially the only known truthful mechanisms in this setting. For project fairness, we consider the mean division as a fair baseline, and bound the maximum difference between the funding received by any project and this baseline. We propose a novel and simple moving phantom mechanism that provides optimal project fairness guarantees. As a corollary of our results, we show that our new mechanism minimizes the L1 distance to the mean for three projects and gives the first non-trivial bounds on this quantity for more than three projects.

Fault diagnosis of lithium-ion batteries based on wavelet packet decomposition and Manhattan average distance

ABSTRACT As lithium-ion batteries are widely used in electric vehicles, safety accidents caused by battery failures emerge one after another. Nevertheless, failures caused by changes in the internal structure or characteristics of the battery, such as sudden and progressive failures, are still a serious problem for electric vehicles, challenging existing fault diagnosis methods. This paper first performs wavelet packet decomposition on the battery’s raw voltage signal to obtain high-quality low-frequency and high-frequency characteristic signal components. Then performs singular value decomposition on the characteristic signal components to extract the corresponding singular value characteristic parameters, and introduces the Manhattan average distance algorithm to battery faults. Diagnosing and locating faulty battery units using the Laida criterion (3-σ criterion) outlier detection method. Finally, actual vehicle data were used to verify the reliability, stability, accuracy of the method, and compared with the traditional Manhattan distance, correlation coefficient, information entropy methods. The method in this paper has good fault detection effects on vehicles with sudden and progressive faults vehicles.

GIS TOOLS IN THE CONSERVATION AND SUSTAINABLE DEVELOPMENT NATIONAL PARKS, FORESTS AND RURAL AREAS

The problem of optimising the location of crops/plots in protected areas, national parks, farmland or forests involves trying to reconcile their proper functioning with nature conservation, the preservation of unique ecosystems or sustainable development. Such areas are crucial for the preservation of biodiversity and wildlife. This paper examines and discusses the impact of land consolidation in rural areas using GIS (Geographic Information System) tools. It performs a series of network analyses to determine the real distances (following the road network) and linear distances (rectilinear distances) between habitats and farmland. A tool was also developed to automate this process. The results obtained were visualised using map compositions, tables and graphs. The implemented project had a significant beneficial effect on the change in the land structure in the analysed village. The number of registered plots decreased by 40% (from 1,189 to 711) due to the consolidation process.

Identification of similarities and clusters of bread baking recipes based on data of ingredients

Abstract We define the similarity of bakery recipes using different distance calculations and identify groups of similar recipes using different clustering algorithms. Our analyses are based on the relative amounts of ingredients included in the recipes. We compare different clustering algorithms (k-means, k-medoid, and hierarchical clustering) to find the optimal number of clusters. Besides the standard distance calculation (euclidean distance), we test three other distance metrics (hamming distance, manhattan distance, and cosine similarity). Additionally, we reduce the impact of raw materials used in large quantities by applying two different data transformations, namely the logarithm of the original data and the binarization of the original data. Clustering recipes based on their ingredients can improve the search for similar recipes and therefore help with the time-consuming process of developing new recipes. Using the hierarchical clustering on the logarithm of the original data, we can separate 704 recipes into three different clusters, achieving a Silhouette Score of 0.531. We visualize our results via dendrograms representing the recipes’ hierarchical separation into individual groups and sub-groups.

Black-Box Boundary Attack Based on Gradient Optimization

Deep neural networks have gained extensive applications in computer vision, demonstrating significant success in fundamental research tasks such as image classification. However, the robustness of these networks faces severe challenges in the presence of adversarial attacks. In real-world scenarios, addressing hard-label attacks often requires the execution of tens of thousands of queries. To combat these challenges, the Black-Box Boundary Attack leveraging Gradient Optimization (GOBA) has been introduced. This method employs a binary search strategy to acquire an initial adversarial example with significant perturbation. The Monte Carlo algorithm is utilized to estimate the gradient of the sample, facilitating iterative movement along the estimated gradient and the direction of the malicious label. Moreover, query vectors positively correlated with the gradient are extracted to construct a sampling space with an optimal scale, thereby enhancing the efficiency of the Monte Carlo algorithm. Experimental evaluations were conducted using the HSJA, QEBA, and NLBA attack methodologies on the ImageNet, CelebA, and MNIST datasets, respectively. The results indicate that, under the constraint of 3 k query times, the GOBA, compared to other methods, can, on average, reduce perturbation (L2 distance) by 55.74% and simultaneously increase the attack success rate by an average of 13.78%.

Collective group drift in a partial-differential-equation-based opinion dynamics model with biased perception kernels.

In the age of technology, individuals accelerate their biased gathering of information, which in turn leads to a population becoming extreme and more polarized. Here we study a partial-differential-equationmodel for opinion dynamics that exhibits collective behavior subject to nonlocal interactions. We developed an interaction kernel function to represent biased information gathering. Through a linear stability analysis, we show that biased populations can still form opinionated groups. However, a population that is too heavily biased can no longer come to a consensus, that is, the initial homogeneous mixed state becomes stable. Numerical simulations with biased information gathering show the ability for groups to collectively drift towards one end of the opinion space. This means that a small bias in each individual will collectively lead to groups of individuals becoming extreme together. The characteristic time scale for a group's existence is captured from numerical experiments using the temporal correlation function. Supplementing this, we included a measure of how different each population is after regular time intervals using a form of the Manhattan and Euclidean distance metrics. We conclude by exploring how wall boundary conditions induce pattern formation initially on the most extreme sides of the domain.

Small codes

AbstractDetermining the maximum number of unit vectors in with no pairwise inner product exceeding is a fundamental problem in geometry and coding theory. In 1955, Rankin resolved this problem for all , and in this paper, we show that the maximum is for all , answering a question of Bukh and Cox. Moreover, the exponent is best possible. As a consequence, we obtain an upper bound on the size of a ‐ary code with block length and distance , which is tight up to the multiplicative factor for any prime power and infinitely many . When , this resolves a conjecture of Tietäväinen from 1980 in a strong form and the exponent is best possible. Finally, using a recently discovered connection to ‐ary codes, we obtain analogous results for set‐coloring Ramsey numbers.

Collaborative Filtering-based Fast Delay-aware algorithm for joint VNF deployment and migration in edge networks

As Network Function Virtualization (NFV) continues to advance, Virtual Network Functions (VNFs) such as firewalls are increasingly used. Service Function Chains (SFCs) are formed by combining specific VNFs in a particular order, which are then deployed on the physical network to provide dedicated services to end users. Occasionally, partial VNF migration is employed to maintain service and network stability. However, decision-making times and system delays may become unacceptable due to the heterogeneous resource requirements of VNFs and the massive state migration of VNFs, especially in Multi-access Edge Computing (MEC) networks where resources are scarce and demand fluctuations are frequent. To solve these challenges, we first formulate the problem of Minimizing decision Time and system Latency for joint VNF Deployment and Migration (MTLDM) as a multi-objective optimization problem. Then, we propose a Collaborative Filtering-based Fast Delay-aware algorithm (CFFD) to solve this problem. In this algorithm, we introduce an innovative approach, referred to as the collaborative filtering-based method, which utilizes the preference information of deployed/migrated VNFs to assist the current VNF deployment/migration in reducing decision-making time. Additionally, we design a similarity-based method in CFFD to search for suitable hosts for the current VNF, thereby reducing the complexity caused by heterogeneity and minimizing system latency. Furthermore, we implement a heuristic method in CFFD to increase the number of accepted requests. In the end, extensive simulations are conducted to evaluate the performance of CFFD in comparison with baseline algorithms and to select the suitable similarity algorithm. The results of the selection simulation show that Manhattan distance and cosine similarity are superior to Pearson’s correlation. Moreover, the comparison simulation results indicate that CFFD outperforms the baseline algorithms in terms of delay optimization and decision time by up to 22.08% and 99%, respectively.

ANALYZING THE EFFECT OF SIMILARITY FUNCTIONS ON PARTITIONING AROUND MEDOIDS ALGORITHM FOR MAPPING DHF DISEASE IN NORTH SUMATRA

Dengue hemorrhagic fever (DHF) is an acute febrile illness caused by a virus through the Aedes mosquito. North Sumatra is among the three provinces with the highest incidence and mortality rates in Indonesia. Mapping of DHF cases is very important in efforts to control and prevent the disease. The Partitioning Around Medoid (PAM) algorithm is commonly used to cluster DHF cases. The idea of PAM is a clustering algorithm with a similarity-based approach to grouping objects in one cluster. There are two main focuses in the research: mapping regencies/cities based on dengue case information and analyzing the performance of several similarity functions. The dataset includes variables of incidence rate (IR), case fatality rate (CFR), larva-free rate (ABJ), and population, obtained from the North Sumatra Provincial Health Office and the Central Statistics Agency (BPS). The analysis showed that three clusters were formed in North Sumatra Province. The first cluster includes regencies/cities such as Langkat, Deli Serdang, Karo, Simalungun, Dairi, Samosir, Humbahas, North Labuhan Batu, North Padang Lawas, South Labuhan Batu, Padang Sidempuan, Nias, South Nias, North Nias, and Sibolga. The second cluster consists of regencies/cities such as Medan, Binjai, Sedang Berdagai, Tebing Tinggi, Batubara, Asahan, Tanjung Balai, Labuhan Batu, Toba, North Tapanuli, Central Tapanuli, Gunungsitoli, and West Nias. The third cluster includes the regencies of South Tapanuli and Mandailing Natal. In addition, an evaluation was conducted using the Silhouette Index to measure the quality of the clustering. Based on the comparison using distance methods (Euclidean distance, Manhattan distance, Minkowski distance, and Chebyshev distance), the highest Silhouette Index value was obtained using Chebyshev distance, which amounted to 0.527554. This value indicates reasonable cluster quality. Thus, this study contributes to the mapping of DHF cases in North Sumatra Province and can be the basis for decision-making in overcoming the disease.

The contribution of suprasegmental cues to perceived L1 and L2 accent distance

First (L1) and second language (L2) pronunciations that diverge more segmentally from the local accent are rated as more distant. However, fewer studies have addressed relations between suprasegmental features, such as intonation, and accent distance. We examined the relation between perceived accent distance and four suprasegmental measures: time normalized f0 Euclidean distance, mean F0, pitch range, and speaking rate. Thirty adults completed six ladder tasks in which listeners ranked talkers’ perceived distance from the local accent. For each ladder, listeners heard one sentence produced by 24 talkers: one man and one woman from six L1 and six L2 English varieties. Each suprasegmental measure independently predicted distance rankings with a variety of significant interactions. Talkers with higher accent rankings had narrower pitch ranges, lower mean f0, faster speech, and greater f0 Euclidean distance. Furthermore, suprasegmental measures related to accent distance differentially based on syntactic structure. For example, as f0 Euclidean distance increased, accent rankings increased for interrogative sentences, but not for declarative sentences. These data support the need for future studies examining how suprasegmental features impact perception of L1 and L2 accents. [Work supported by NSF grants 1941691 and 1941662].

A Region-Based Approach for the Operational Reliability Evaluation of Power Systems With Renewable Energy Integration

Power system operational reliability evaluation is computationally time-consuming due to the need to analyze numerous system states resulting from uncertain renewable energies and loads as well as random outages of components. This paper presents a load feasible region (LFR) approach to accelerate the process of operational reliability evaluation. The LFR of each state, which is defined as a polyhedron in the load space, can be built in advance offline. The inner points of the LFR correspond to the loads that can be satisfied under the system operational constraints. Then, the preestablished LFR can be corrected with little computing effort considering the renewable energy fluctuation during the operation phase. The optimal load shedding under each state can be obtained directly by calculating the minimum Manhattan distance to the LFR's boundary. Multiple application examples are presented to demonstrate the efficiency and scalability of the proposed region-based approach.

Swim start and performance in 50 m freestyle in different age categories of competitive swimmers

Background and Study Aim. In international races, the winners are decided by hundredths of a second, which is why the swim start plays an important role, especially in the sprint disciplines. The aim of the study is to reveal the differences in kinematic parameters of start and performance in the sprint 50 m freestyle discipline based on gender in different age categories of competitive swimmers at international competitions organized in Slovakia. Material and Methods. The sample consisted of 180 females and 189 males who were divided into age categories (K1, K2, K3). SwimPro cameras and Dartgish software were used to monitor kinematic parameters. The parameters monitored were - block time (BT), time (FT) and distance (FD) of flight, time (UWT) and distance (UWD) underwater, time to 15 m (T15), 25 m (T25) and 50 m (T50). Data were tested by Shapiro-Wilk, Kurskal-Wallis ANOVA and Mann-Whitney U test in Statistica 13.5. Results. In the phase above water level, there were greater differences (p<0.01) in females than in males. Inter-sex differences (p<0.01) were evident in FT in K3, K2 and in FD across all categories. In the underwater phase, differences (p<0.01) were evident in both sexes. Inter-sex differences were more evident in UWT (p<0.01) than UWD (p<0.05). There were inter-sex differences (p<0.01) in ST and SD between all categories except K3. At T15, T25 and T50, differences (p<0.01) were most pronounced between K3 and K2, K1 in females and between all categories in males. Inter-sex differences (p<0.01) were also evident across all categories. Conclusions. The study highlighted differences in 50m freestyle start and performance between age groups and gender, so coaches are advised to design training sessions for swimmers separately.

A Low-Power Analog Integrated Euclidean Distance Radial Basis Function Classifier

This study introduces a low-power analog integrated Euclidean distance radial basis function classifier. The high-level architecture is composed of several Manhattan distance circuits in connection with a current comparator circuit. Notably, each implementation was designed with modularity and scalability in mind, effectively accommodating variations in the classification parameters. The proposed classifier’s operational principles are meticulously detailed, tailored for low-power, low-voltage, and fully tunable implementations, specifically targeting biomedical applications. This design methodology materialized within a 90 nm CMOS process, utilizing the Cadence IC Suite for the comprehensive management of both the schematic and layout design aspects. During the verification phase, post-layout simulation results were meticulously cross-referenced with software-based classifier implementations. Also, a comparison study with related analog classifiers is provided. Through the simulation results and comparative study, the design architecture’s accuracy and sensitivity were effectively validated and confirmed.