按方法、领域与证据探索科学。

Variant Calling

Variant calling is the computational process of identifying positions in a sequenced genome that differ from a reference sequence — including single nucleotide polymorphisms (SNPs), small insertions and deletions (indels), and structural variants. It transforms aligned sequencing reads into an interpretable catalogue o

deep learning

Variational Autoencoder

The Variational Autoencoder (VAE) is a deep generative latent-variable model, introduced by Diederik Kingma and Max Welling in 2014, that encodes data as a probability distribution in a latent space and samples from that distribution to generate new examples. It is used for data generation, anomaly detection, and featu

signal processing

Variational Mode Decomposition

Variational Mode Decomposition (VMD) is a fully adaptive, non-recursive signal decomposition method introduced by Konstantin Dragomiretskiy and Dominique Zosso in 2014. It decomposes a real-valued input signal into a discrete number of sub-signals, called intrinsic mode functions (IMFs), each with a specific sparsity i

quantum computing

Variational Quantum Eigensolver

The Variational Quantum Eigensolver (VQE) is a hybrid quantum-classical algorithm designed to find the lowest eigenvalue (ground state energy) of a quantum Hamiltonian. Introduced by Peruzzo et al. in 2014, it exploits the variational principle to combine the power of quantum circuits with classical optimization to sol

particle physics

Vegas Monte Carlo

VEGAS is an adaptive Monte Carlo algorithm for numerical integration of multidimensional functions, particularly useful for high-dimensional integrals common in particle physics calculations. By adaptively refining the sampling distribution to concentrate points in high-contribution regions, VEGAS dramatically improves

meteorology

Velocity-Azimuth Display

The Velocity-Azimuth Display (VAD) is a radar analysis technique that extracts the radial velocity of wind at a constant altitude as a function of azimuth angle around the radar. By fitting a sinusoidal pattern to these measurements, VAD retrieves the mean wind speed and direction at that altitude, providing wind profi

deep learning

VGGNet

VGGNet is a deep convolutional neural network architecture introduced by Karen Simonyan and Andrew Zisserman at the Visual Geometry Group, Oxford, in 2014 (published at ICLR 2015). It demonstrated that network depth — achieved exclusively through stacking small 3x3 convolutional filters — is the single most critical fa

materials science

Vickers Hardness

Vickers Hardness testing is a mechanical characterization technique for determining material hardness by pressing a diamond pyramid indenter into a material surface under controlled load and measuring the resulting indent dimensions. Invented by Smith and Sandland in 1922, Vickers hardness is applicable across an enorm

deep learning

Vision Mamba

Vision Mamba is an efficient state space model approach for image understanding introduced in 2024 that adapts Mamba, a linear-complexity sequence model, to computer vision. By reformulating image tokens as sequences and using state space models, Vision Mamba achieves competitive accuracy with transformers while mainta

deep learning

Vision Transformer

The Vision Transformer (ViT), introduced by Dosovitskiy and colleagues in 2021, splits an image into fixed-size patches, treats those patches as a sequence, and applies the Transformer self-attention mechanism to image classification. Given enough training data, it surpasses convolutional neural networks (CNNs).

deep learning

Visual Contrastive Learning

Visual contrastive learning is a self-supervised deep-learning approach — popularised by frameworks such as SimCLR (Chen et al., 2020) and MoCo (He et al., 2020) — that learns rich image representations without labels by pulling different augmentations of the same image together and pushing different images apart. It t

music information retrieval

Vocal Separation

Vocal separation is the task of isolating the singing voice from a mixed music recording, leaving the instrumental accompaniment. Introduced formally by Han et al. (2012), it is critical for music editing, remixing, karaoke generation, and music analysis. Modern deep learning approaches (Défossez et al., 2021) have ach

fluid dynamics

Volume of Fluid

The Volume of Fluid (VOF) method is an Eulerian interface-tracking technique for multiphase flow simulations, developed by Hirt and Nichols in 1981. Instead of explicitly tracking the interface between phases, VOF advects a scalar field (the volume fraction) that represents the fractional volume of one phase in each gr

machine learning

Voting Ensemble

A voting ensemble trains several diverse classifiers independently and combines their predictions by a vote: hard voting picks the class chosen by the most models, while soft voting averages their class-probability estimates, optionally with per-model weights. The combination usually outperforms any individual member,

cryptography

Vulnerability Assessment

Vulnerability assessment is a systematic process of identifying, quantifying, and prioritizing security weaknesses in systems, networks, and applications. Using automated scanning tools and manual techniques, organizations discover unpatched software, misconfigurations, weak cryptographic practices, and other exposures

mining engineering

Washability

Washability analysis is a laboratory method that determines the feasibility and efficiency of density-based separation for coal or mineral beneficiation. By fractionating ore or coal into density bins using sink-float tests and assaying each fraction, engineers can optimize design of separation plants (dense-medium cyc

deep learning

Wasserstein GAN

Wasserstein GAN (WGAN) is a generative adversarial network variant introduced by Arjovsky, Chintala, and Bottou in 2017 that replaces the Jensen-Shannon divergence used in the original GAN with the Wasserstein-1 (Earth Mover) distance. This substitution provides a theoretically grounded training objective that yields m

computer vision

Watershed Segmentation

Watershed segmentation is a morphological image processing technique that automatically segments an image into distinct regions by treating image intensity as a topographic landscape where each object corresponds to a valley. Introduced by Beucher and Lantuéjoul in 1979 and refined by Meyer, the watershed algorithm is

finance

Wavelet Financial Analysis

Wavelet financial analysis decomposes a financial time series into different frequency bands (time scales) so short- and long-term relationships can be studied at the same time. Drawing on the treatments of Gençay, Selçuk and Whitcher (2001) and Aguiar-Conraria and Soares (2014), wavelet coherence then visualises how t

time series

Wavelet Neural Network

A wavelet neural network (WNN) is a function approximation architecture that uses wavelet functions as activation functions in place of traditional sigmoid or ReLU functions. Introduced by Zhang and Benveniste (1992), WNNs combine the multiscale decomposition properties of wavelets with the learning capabilities of neu

architecture

Wayfinding Analysis

Wayfinding Analysis is a method for assessing how easily people can navigate and orient themselves in buildings and urban environments. Rooted in Kevin Lynch's concept of legibility and developed further by Romedi Passini, it combines cognitive psychology, design principles, and empirical testing to diagnose navigation

astronomy

Weak Gravitational Lensing

Weak gravitational lensing occurs when light from distant sources bends slightly as it travels through the universe, passing through the gravitational fields of matter concentrations. Proposed theoretically by Nick Kaiser in 1992, this subtle effect has become one of the most powerful cosmological probes, directly reve

deep learning

Weakly supervised BERT-based classification

Weakly supervised BERT-based classification adapts BERT to text classification tasks when only noisy, heuristic, or programmatically generated labels are available instead of clean human annotations. It combines weak supervision frameworks — such as labeling functions and data programming — with BERT's pre-trained lang

deep learning

Weakly supervised convolutional neural network

A weakly supervised CNN is a convolutional neural network trained with incomplete, coarse, or noisy annotations instead of full pixel-level or bounding-box labels. Typical weak labels include image-level class tags, partial annotations, or crowd-sourced noisy labels. The model learns to classify and often to roughly lo

deep learning

Weakly Supervised Diffusion Model

A weakly supervised diffusion model trains or conditions a denoising diffusion probabilistic model using coarse, noisy, or incomplete supervision signals — such as image-level class labels, bounding boxes, or crowd-sourced annotations — instead of pixel-precise ground truth. This allows high-quality generative and disc

deep learning

Weakly supervised GAN

A Weakly Supervised GAN is a generative adversarial network trained with partially labeled, noisily labeled, or coarse-annotation data instead of fully annotated ground truth. It extends the standard GAN framework so that limited supervision guides conditional generation or discriminative learning, enabling high-qualit

deep learning

Weakly supervised graph neural network

A Weakly Supervised Graph Neural Network (WS-GNN) is a graph deep-learning approach that learns from graph-structured data — nodes, edges, and their attributes — when only noisy, partial, or indirectly obtained labels are available. By coupling GNN message passing with noise-robust training strategies, it extends graph

deep learning

Weakly Supervised GRU

Weakly Supervised GRU trains a Gated Recurrent Unit network on sequences labeled by imperfect, heuristic, or programmatic sources rather than costly hand-annotated ground truth. It combines the GRU's efficiency at capturing temporal dependencies with weak-supervision techniques that aggregate noisy labels, enabling pra

deep learning

Weakly Supervised Image Classification

Weakly supervised image classification trains convolutional or transformer-based networks using only coarse, incomplete, or noisy supervision — such as image-level category labels, hashtags, or web-scraped tags — without requiring precise bounding boxes or pixel annotations. This dramatically reduces labeling cost whil

deep learning

Weakly Supervised Instance Segmentation

Weakly supervised instance segmentation trains deep networks to delineate individual object instances at pixel level using only cheap, incomplete annotations — such as bounding boxes, image-level labels, or point clicks — rather than costly full pixel-wise masks. It dramatically reduces annotation effort while still pr

deep learning

Weakly supervised LDA topic model

Weakly Supervised LDA is an extension of Latent Dirichlet Allocation that incorporates lightweight human guidance — typically keyword seeds or must-link/cannot-link constraints — into the Dirichlet priors, steering learned topics toward domain-meaningful themes without requiring fully labeled documents. It sits between

deep learning

Weakly supervised LSTM

Weakly supervised LSTM trains a Long Short-Term Memory network on sequence data where clean, manually annotated labels are scarce or absent. Instead, multiple imperfect label sources — heuristic rules, distant supervision, crowdsourcing, or programmatic labeling functions — are combined to produce probabilistic trainin

deep learning

Weakly supervised multilayer perceptron

A Weakly Supervised Multilayer Perceptron trains a standard feedforward neural network when only imperfect supervision is available — labels may be noisy, incomplete, crowd-sourced, rule-generated, or derived from distant supervision — enabling learning at scale without the cost of full expert annotation.

deep learning

Weakly Supervised Object Detection

Weakly Supervised Object Detection (WSOD) trains object detectors using only image-level labels — indicating which object classes appear in an image — without requiring costly bounding-box annotations. Multiple Instance Learning (MIL) formulations allow the model to discover the likely location of each object class fro

deep learning

Weakly supervised question answering

Weakly supervised question answering (WS-QA) trains neural reading-comprehension models using indirect or automatically derived answer labels rather than expensive human-annotated span annotations. By exploiting distant supervision, heuristic labeling, or answer-presence signals, WS-QA makes QA feasible in domains and

deep learning

Weakly supervised recurrent neural network

A weakly supervised RNN trains a recurrent neural network on sequences whose labels come from imperfect sources — heuristic rules, distant supervision, crowdsourcing, or generative label models — rather than expensive expert annotation. This lets researchers exploit large unlabeled corpora for sequential tasks such as

deep learning

Weakly supervised reinforcement learning

Weakly supervised reinforcement learning (WSRL) trains agents in environments where the reward signal is imperfect, sparse, delayed, or only partially informative — unlike dense fully-supervised RL. The agent must learn effective policies despite incomplete feedback, using auxiliary signals, reward modeling, or prefere

deep learning

Weakly Supervised RoBERTa-based Classification

Weakly supervised RoBERTa-based classification combines the RoBERTa pretrained transformer with weak supervision — programmatic or heuristic labeling sources — to train powerful text classifiers without requiring a fully hand-labeled dataset. Labeling functions, distant supervision, or crowd-sourced signals generate no

deep learning

Weakly Supervised Semantic Segmentation

Weakly Supervised Semantic Segmentation (WSSS) trains pixel-level scene parsers using only cheap, coarse annotations — typically image-level class tags — instead of costly dense pixel masks. By generating proxy pseudo-labels from a classification network (via Class Activation Maps or similar localisation cues) and iter

deep learning

Weakly supervised sentence embeddings

Weakly supervised sentence embeddings train dense sentence representations using noisy, heuristic, or programmatically generated labels instead of costly human annotation. Labeling functions — rules, distant supervision signals, or lightweight classifiers — supply approximate supervision that a label model aggregates i

deep learning

Weakly supervised text summarization

Weakly supervised text summarization trains abstractive or extractive summarization models without manually annotated reference summaries. Instead of costly human labels, it exploits weak signals — heuristic rules, distant supervision, noisy automatic labels, or self-supervised objectives — to guide sequence-to-sequenc

deep learning

Weakly Supervised Topic Modeling

Weakly supervised topic modeling incorporates lightweight domain knowledge — typically seed words or soft constraints — into a probabilistic topic model to steer discovered topics toward researcher-meaningful themes. It sits between fully unsupervised LDA and supervised classifiers, requiring far less annotation than t

deep learning

Weakly supervised transformer

Weakly Supervised Transformer combines the representational power of Transformer architectures with weak supervision strategies that exploit noisy, incomplete, or programmatically generated labels — making it possible to train high-quality NLP and vision models when fully annotated datasets are scarce or prohibitively

deep learning

Weakly Supervised Variational Autoencoder

A Weakly Supervised Variational Autoencoder (WS-VAE) extends the standard VAE generative framework by incorporating partial, noisy, or coarse supervision signals — such as crowd-sourced labels, heuristic rules, or programmatic annotations — to guide latent space learning without requiring fully annotated data. It is wi

deep learning

Weakly supervised vision transformer

Weakly Supervised Vision Transformer (WS-ViT) trains a Vision Transformer on image data that lacks precise pixel-level annotations, instead using cheaper, noisier supervision such as image-level class tags, bounding boxes, or web-scraped text. The global self-attention mechanism of the transformer makes it especially c

deep learning

Weakly supervised Word2Vec

Weakly Supervised Word2Vec trains Word2Vec-style embeddings using automatically generated, noisy, or heuristic labels rather than costly manual annotation. By leveraging labeling functions, distant supervision, or keyword-based rules to assign soft labels, the approach enables domain-adapted word representations even w

aerospace

Weight and Balance

Weight and balance analysis is the process of determining the total weight of an aircraft and the location of its center of gravity (CG) throughout its operational envelope. Essential for aircraft safety and performance, weight and balance ensures that the CG remains within allowable limits (forward and aft) to maintai

network analysis

Weighted Betweenness Centrality

Weighted Betweenness Centrality extends Freeman's betweenness measure to edge-weighted graphs by routing shortest paths through a tunable transformation of edge weights. Nodes that sit on many high-value shortest paths receive high scores, identifying brokers and bridges in social, biological, and information networks

network analysis

Weighted Closeness Centrality

Weighted closeness centrality extends the classic closeness measure to networks where edges carry numerical weights — such as frequency, strength, or cost — by incorporating those weights into shortest-path distances. Nodes that can reach others quickly along strong or efficient connections receive higher scores, makin

network analysis

Weighted Community Detection

Weighted community detection identifies densely connected groups — communities — in networks where edges carry numeric strengths (weights). By incorporating edge weights into the modularity function, it reveals structure that binary adjacency alone would miss: two nodes connected by a strong tie are treated as more sim

network analysis

Weighted Degree Centrality

Weighted degree centrality — also called node strength — extends the classic degree centrality measure to networks whose edges carry numeric weights. Instead of simply counting a node's connections, it sums the weights of all edges incident to that node, capturing both the volume and the intensity of a node's ties in a

network analysis

Weighted Ego Network Analysis

Weighted ego network analysis examines the personal network of a focal actor (the ego) and incorporates tie strength — measured as interaction frequency, closeness, or resource exchange — as edge weights. By moving beyond simple presence or absence of a tie, it captures how much each relationship matters and how those

network analysis

Weighted Eigenvector Centrality

Weighted eigenvector centrality extends the classic eigenvector centrality measure to graphs where edges carry numerical weights, scoring each node proportionally to the sum of its neighbors' scores multiplied by the connecting edge weights. Nodes score highly not just by having many connections but by being strongly l

network analysis

Weighted Exponential Random Graph Model

The Weighted Exponential Random Graph Model (W-ERGM) extends the classic binary ERGM framework to networks whose edges carry quantitative values — such as frequency of contact, trade volume, or collaboration intensity. It models the entire valued-edge network as a probability distribution defined over all possible weig

model evaluation

Weighted F1

Weighted F1 computes the F1-score for each class and then takes a weighted average, where weights are proportional to the number of samples in each class (support). It provides a middle ground between macro and micro-averaging.

network analysis

Weighted Knowledge Graph Analysis

Weighted Knowledge Graph Analysis extends standard knowledge graph methods by assigning numerical weights — such as confidence scores, co-occurrence frequencies, or relation strengths — to edges between entities. These weights allow analysts to prioritise high-confidence triples, find the most influential paths, and co

network analysis

Weighted Modularity Analysis

Weighted modularity analysis extends the classical Newman-Girvan modularity measure to networks where edges carry numeric strengths (frequencies, intensities, costs). By replacing binary adjacency with tie weights, it finds community partitions that reflect how densely interconnected subgroups are relative to what is e

network analysis

Weighted Multiplex Network Analysis

Weighted multiplex network analysis studies systems in which the same set of actors are connected through multiple types of relationships simultaneously, and each relationship carries a quantitative strength or frequency. By capturing both the variety and the intensity of ties across layers, it reveals patterns invisib

network analysis

Weighted Network Diffusion Analysis

Weighted Network Diffusion Analysis models how information, influence, disease, or resources spread through a network whose edges carry quantitative strength values. By letting tie weights govern transition probabilities, the method produces more realistic spreading dynamics than binary-edge diffusion, revealing which

network analysis

Weighted PageRank

Weighted PageRank extends the classic PageRank algorithm to networks where edges carry different strengths or frequencies, distributing importance proportionally to both incoming and outgoing edge weights rather than treating all links equally. This makes it substantially more informative than binary PageRank in any ne