Interpretability in machine learning (ML) is crucial for high stakes decisions and troubleshooting. In this work, we provide fundamental principles interpretable ML, dispel common misunderstandings that dilute the importance of topic. We also identify 10 technical challenge areas history background on each problem. Some these problems are classically important, some recent have arisen last few years. These are: (1) Optimizing sparse logical models such as decision trees; (2) Optimization...

Dimension reduction (DR) techniques such as t-SNE, UMAP, and TriMAP have demonstrated impressive visualization performance on many real world datasets. One tension that has always faced these methods is the trade-off between preservation of global structure local structure: can either handle one or other, but not both. In this work, our main goal to understand what aspects DR are important for preserving both it difficult design a better method without true understanding choices we make in...

Abstract Dimension reduction (DR) algorithms project data from high dimensions to lower enable visualization of interesting high-dimensional structure. DR are widely used for analysis single-cell transcriptomic data. Despite widespread use such as t-SNE and UMAP, these have characteristics that lead lack trust: they do not preserve important aspects structure sensitive arbitrary user choices. Given the importance gaining insights DR, methods should be evaluated carefully before trusting...

Dimension reduction (DR) algorithms have proven to be extremely useful for gaining insight into large-scale high-dimensional datasets, particularly finding clusters in transcriptomic data. The initial phase of these DR methods often involves converting the original data a graph. In this graph, each edge represents similarity or dissimilarity between pairs points. However, graph is frequently suboptimal due unreliable distances and limited information extracted from This problem exacerbated...

Point cloud completion aims to reconstruct complete 3D shapes from partial point clouds. With advancements in deep learning techniques, various methods for have been developed. Despite achieving encouraging results, a significant issue remains: these often overlook the variability clouds sampled single object surface. This can lead ambiguity and hinder achievement of more precise results. Therefore, this study, we introduce novel network, namely Dual-Codebook Completion Network (DC-PCN),...

Point cloud completion aims to reconstruct complete 3D shapes from partial point clouds. With advancements in deep learning techniques, various methods for have been developed. Despite achieving encouraging results, a significant issue remains: these often overlook the variability clouds sampled single object surface. This can lead ambiguity and hinder achievement of more precise results. Therefore, this study, we introduce novel network, namely Dual-Codebook Completion Network (DC-PCN),...

Large language models (LLMs) have significantly facilitated human life, and prompt engineering has improved the efficiency of these models. However, recent years witnessed a rise in engineering-empowered attacks, leading to issues such as privacy leaks, increased latency, system resource wastage. Though safety fine-tuning based methods with Reinforcement Learning from Human Feedback (RLHF) are proposed align LLMs, existing security mechanisms fail cope fickle highlighting necessity...

To quantitatively evaluate chronic kidney disease (CKD), a deep convolutional neural network-based segmentation model was applied to renal enhanced computed tomography (CT) images. A retrospective analysis conducted on cohort of 100 individuals diagnosed with CKD and 90 healthy kidneys, who underwent contrast-enhanced CT scans the kidneys or abdomen. Demographic clinical data were collected from all participants. The study consisted two distinct stages: firstly, development validation...

From the viewpoint of vibration control, if amplitude main frequencies response can be reduced, energy structure is greatly reduced. Modal parameters, including modal shapes, natural frequencies, and damping ratios, reflect dynamics used to control vibration. This paper integrates idea “forgetting factor weighting” into eigenvector recursive principal component analysis, then proposes an operational analysis (OMA) method that uses PCA with a forgetting (ERPCAWF). The proposed identify...

A large number of smart devices make the Internet Things world smarter. However, currently cloud computing cannot satisfy real-time requirements and fog is a promising technique for processing. Operational modal analysis obtains parameters that reflect dynamic properties structure from vibration response signals. In Things, operational method can be embedded in to achieve structural health monitoring fault detection. this article, four-layer framework combining designed. This introduces...

Interpretability in machine learning (ML) is crucial for high stakes decisions and troubleshooting. In this work, we provide fundamental principles interpretable ML, dispel common misunderstandings that dilute the importance of topic. We also identify 10 technical challenge areas history background on each problem. Some these problems are classically important, some recent have arisen last few years. These are: (1) Optimizing sparse logical models such as decision trees; (2) Optimization...

In order to identify the modal parameters of time invariant three-dimensional engineering structures with damping and small nonlinearity, a novel isometric feature mapping (Isomap)-based operational analysis (OMA) method is proposed extract nonlinear features in this paper. Using Isomap-based OMA method, low-dimensional embedding matrix multiplied by transformation obtain original matrix. We find correspondence relationships between coordinate response shapes. From matrix, natural...

Visual concept discovery has long been deemed important to improve interpretability of neural networks, because a bank semantically meaningful concepts would provide us with starting point for building machine learning models that exhibit intelligible reasoning process. Previous methods have disadvantages: either they rely on labelled support sets incorporate human biases objects are "useful," or fail identify multiple occur within single image. We reframe the task as an unsupervised...

To predict the multi-point vibration response in frequency domain when uncorrelated multi-source loads are unknown, a data-driven and multi-input multi-output least squares support vector regression (MIMO LS-SVR)-based method is proposed. Firstly, relationship between measured unmeasured formulated using transfer function domain. Secondly, multiple analysis problem of prediction described formally, its mathematical model established. With as input output, history data assembled MIMO training...

Parametric dimensionality reduction methods have gained prominence for their ability to generalize unseen datasets, an advantage that traditional approaches typically lack. Despite growing popularity, there remains a prevalent misconception among practitioners about the equivalence in performance between parametric and non-parametric methods. Here, we show these are not equivalent -- retain global structure but lose significant local details. To explain this, provide evidence parameterized...

Dimension reduction (DR) algorithms have proven to be extremely useful for gaining insight into large-scale high-dimensional datasets, particularly finding clusters in transcriptomic data. The initial phase of these DR methods often involves converting the original data a graph. In this graph, each edge represents similarity or dissimilarity between pairs points. However, graph is frequently suboptimal due unreliable distances and limited information extracted from This problem exacerbated...

To address the problems of singularities, sensitivity to measurement noise, and low efficiency in traditional principal component analysis (PCA)-based operational modal (OMA), we present a Sanger neural network (SNNPCA) algorithm identify mod al parameters. SNNPCA is two-layer that trained using generalized Hebbian ensure its output converges components. After has converged, link weights correspond separation matrix PCA. In SNNPCA-based OMA, response points are set as input neurons,...

In order to select the window function and size adaptively before getting results, we proposed adaptive moving principle component analysis (AMWPCA) based OMA method identify modal shapes natural frequencies of slow LTV structures with weekly damped only from non-stationary vibration response signal online. The is achieved in two ways: change or size. We develop an indicator as basis for changes. Our approach make difference between adjacent eigenvalues not too small. operational parameter...