연구 방법을 한데 모은 단 하나의 카탈로그 — 각 방법이 어떻게 작동하는지, 언제 쓰는지, 무엇을 할 수 없는지 알아보세요.
Differential variant calling is a bioinformatics pipeline that identifies genetic variants — single nucleotide variants (SNVs), small insertions/deletions (indels), and structural variants — that are present in one biological sample or condition but absent (or significantly enriched) in a paired reference sample. The c
The Diffie-Hellman key exchange, invented by Whitfield Diffie and Martin Hellman in 1976, is a foundational protocol for establishing a shared secret over an insecure communication channel. Two parties who have never previously communicated can use Diffie-Hellman to agree on a symmetric encryption key that an eavesdrop
DiffServ is a QoS architecture providing scalable, class-based service differentiation in networks. Introduced by IETF (1998), DiffServ marks packets with a Differentiated Services Code Point (DSCP) in the IP header, enabling routers to apply per-hop-behaviors (PHBs) based on markings. Unlike IntServ (which reserves re
A diffusion model is a generative deep-learning method, introduced by Ho, Jain and Abbeel in 2020 (DDPM), that learns to produce high-quality images, audio and molecular structures by reversing a step-by-step noising process. It has largely displaced GANs as the current state of the art in generative modelling.
A digital signature scheme provides authentication, integrity assurance, and non-repudiation of electronically signed documents. Using public-key cryptography (such as RSA, DSA, or ECDSA), the originator signs a message with a private key in a way that any recipient can verify the signature using the originator's publi
Digital Twin Simulation, first conceptualised by Michael Grieves at the University of Michigan around 2002 and formally described in his 2014 white paper, creates a continuously updated virtual copy of a physical system by fusing real-time sensor data with a mechanistic (physics-based) model and machine-learning compon
A Dilated CNN is a one-dimensional convolutional network whose receptive field grows exponentially with depth, letting it model long-range structure in time series and audio signals. WaveNet (van den Oord et al., 2016) and the Temporal Convolutional Network of Bai, Kolter and Koltun (2018) are the prominent members of
Direct Numerical Simulation (DNS) is a computational approach that solves the Navier-Stokes equations without turbulence models, resolving all scales of motion from the largest energy-containing eddies down to the smallest dissipative scales (Kolmogorov microscales). Pioneered by Steven Orszag in 1971, DNS provides com
Direct Preference Optimization (DPO) is a training method introduced by Rafailov et al. in 2023 that aligns language models with human preferences without requiring an explicit reward model. By directly optimizing for preference pairs (better response vs worse response), DPO simplifies the training pipeline compared to
Direct Torque Control (DTC) is a method for controlling induction motors by directly manipulating magnetic flux and torque through switching of power converter inverter arms. Introduced by Takahashi and Noguchi in 1986, DTC provides fast torque response, low harmonic distortion, and robust performance without requiring
Directed Betweenness Centrality extends Freeman's classic betweenness measure to directed graphs, quantifying how often a node lies on the shortest directed paths between all other pairs of nodes. It identifies gatekeepers, brokers, and bottlenecks in asymmetric flows such as information cascades, citation networks, an
Directed closeness centrality extends the classical closeness measure to directed networks by separately quantifying how quickly a node can be reached by others (in-closeness) and how quickly it can reach all others (out-closeness). It is a foundational node-level metric in social network analysis and graph theory, use
Directed community detection identifies densely interconnected groups of nodes in a directed network, accounting for the asymmetry of edges (e.g., A follows B does not imply B follows A). Adapting modularity or flow-based criteria to directed graphs reveals clusters that undirected methods systematically miss, making i
Directed ego network analysis examines the personal network of a focal node — the ego — by distinguishing the direction of each tie: who sends resources, support, or information to the ego, and to whom the ego sends them. This asymmetric perspective reveals role differentiation, dependence, and brokerage that undirecte
Directed eigenvector centrality extends the classic eigenvector centrality to directed graphs by scoring each node according to the centrality of the nodes that point to it (in-direction) or that it points to (out-direction). A node earns a high score not merely by having many connections but by being connected to othe
The Directed Exponential Random Graph Model (Directed ERGM) is a family of statistical models for directed networks that estimates the probability of observing a given directed graph as a function of structural configurations — such as reciprocity, transitive triads, and in-degree centralization — and node or dyad cova
Directed Knowledge Graph Analysis represents factual knowledge as a directed labeled multigraph of entities (nodes) and typed relations (directed edges), enabling structured reasoning, inference, and discovery over large heterogeneous datasets. The direction of edges encodes asymmetric relationships such as 'authored-b
Directed modularity analysis extends the classic Newman-Girvan modularity framework to directed graphs, where edges carry a source and a destination. Formalized by Leicht and Newman in 2008, it partitions nodes into communities by maximizing a modularity score that accounts for each node's separate in-degree and out-de
Directed multiplex network analysis models systems where the same set of nodes are connected by multiple types of directed (asymmetric) relationships across distinct layers — such as citation flows, information cascades, or authority hierarchies co-existing simultaneously. It extends multiplex network analysis by prese
Directed network diffusion analysis studies how information, disease, behavior, or influence spreads through a network in which edges carry direction — meaning transmission flows one way along each link. It combines graph-theoretic representations with stochastic spreading models such as independent cascade, linear thr
Directed PageRank is a link-based authority scoring algorithm that assigns importance scores to nodes in a directed graph by iteratively redistributing rank through outgoing edges. Introduced by Brin and Page in 1998 as the backbone of Google Search, it measures not just how many in-links a node has but how authoritati
Directed Social Network Analysis (directed SNA) studies networks in which every tie has an explicit direction — from a sender to a receiver — rather than treating relationships as symmetric. It extends the classical SNA toolkit with in-degree, out-degree, reciprocity, and asymmetric path measures, making it the appropr
Directed two-mode network analysis studies bipartite graphs in which nodes belong to two distinct sets — such as actors and events, authors and papers, or firms and markets — and edges carry a direction, capturing asymmetric relationships like citation, referral, or endorsement. Combining the duality of two-mode struct
Disclosure Risk Assessment is a probabilistic framework introduced by Duncan and Lambert (1989) for quantifying how likely it is that releasing microdata — individual-level records from surveys or administrative files — will allow an outside party to identify a specific respondent or infer sensitive attributes. It is u
Discourse parsing is a natural-language-processing task that models the rhetorical relations between sentences and paragraphs of a text — relations such as cause, contrast, and elaboration — and represents them as a tree structure. It works within established frameworks, principally Rhetorical Structure Theory (RST), i
Discrete choice simulation is a behavioural modelling method — grounded in random utility theory formalised by Daniel McFadden in the 1970s and extended to simulation-based estimation by Kenneth Train — that estimates how individuals choose among mutually exclusive alternatives and then uses those estimated preference
The discrete wavelet transform (DWT) is a fast, computationally efficient method for decomposing signals into different frequency and time components using orthogonal or biorthogonal wavelet functions. Developed rigorously by Ingrid Daubechies (1992) and built on Mallat's multiresolution decomposition theory (1989), th
Discrete-Event Simulation (DES) is a computational modeling paradigm in which the state of a system changes only at a countable sequence of points in time — the events. Between events nothing changes, so the simulation clock jumps directly from one event to the next. Formalized through the foundational textbooks of Ban
Discrete-event system simulation (DES) is a computational modelling technique in which the state of a system changes only at discrete points in time — called events — such as a customer arriving, a machine starting, or a job completing. Formalised through foundational texts by Kelton, Sadowski, and Zupick (2014) and La
DLinear is a lightweight time series forecasting model introduced by Zeng et al. at AAAI 2023. It challenges the prevailing assumption that Transformer-based architectures are necessary for accurate long-horizon forecasting. The model decomposes an input sequence into trend and seasonal components using a moving averag
Doc2Vec, also known as Paragraph Vector, is a representation-learning method introduced by Le and Mikolov (2014) that maps whole documents to fixed-length dense vectors. These vectors place similar documents close together in space, supporting document comparison and classification.
Document clustering is an unsupervised text-mining task that groups documents with similar content together without using any labels. It is used to organise large collections and for exploratory analysis, drawing on the body of text-mining techniques consolidated by Aggarwal and Zhai (2012) and compared empirically by
Domain adaptation is a natural-language-processing technique that takes a general pretrained language model and fine-tunes it on target-domain data so that it performs better in specialised fields such as medicine, law, and finance. It builds on the transfer-learning ideas behind work like Blitzer et al. (2007) on cros
A domain-adaptive CNN trains a convolutional network on a labeled source domain and adapts its learned feature representations to an unlabeled or lightly labeled target domain, bridging the distribution gap so that visual classifiers transfer reliably across datasets, sensors, or imaging conditions without full re-anno
Domain-adaptive Doc2Vec adapts the Paragraph Vector (Doc2Vec) framework so that document embeddings learned on a source domain transfer effectively to a target domain. By aligning the representation space across domains during or after training, the model produces embeddings that are informative on both, enabling cross
A Domain-Adaptive GAN combines generative adversarial learning with domain adaptation to bridge the distribution gap between a labeled source domain and an unlabeled or sparsely labeled target domain. By training a generator and discriminator adversarially, the model learns domain-invariant representations or translate
Domain-adaptive image classification trains a visual classifier on a labeled source domain and adapts it to a target domain where labeled data are scarce or absent. By aligning feature distributions across domains, the model retains discriminative accuracy on the target distribution without requiring full target re-ann
Domain-adaptive instance segmentation extends Mask R-CNN-style architectures to operate across distribution shifts — training on a labeled source domain (e.g., synthetic renderings or daytime images) and adapting to an unlabeled or weakly labeled target domain (e.g., real scenes or nighttime footage). Adversarial featu
A domain-adaptive multilayer perceptron (DA-MLP) is a feedforward neural network trained to learn representations that are useful across a labeled source domain and an unlabeled or differently distributed target domain. By minimizing both a task loss and a domain-discrepancy objective, the MLP generalizes to the target
Domain-adaptive Named Entity Recognition (DA-NER) applies named entity recognition to a target domain by transferring or adapting a model trained on a source domain, using techniques such as domain-specific pre-training, adversarial alignment, or feature augmentation. It addresses the performance collapse that standard
Domain-adaptive NMF Topic Modeling applies Non-negative Matrix Factorization to discover latent topics across text from multiple domains, using regularization or shared basis constraints to transfer topic knowledge from a resource-rich source domain to a target domain with limited labeled data. It combines interpretabl
Domain-adaptive Question Answering (DA-QA) adapts a pre-trained language model — typically BERT or RoBERTa — first trained on general QA benchmarks such as SQuAD to answer questions accurately in a new target domain (e.g., biomedical, legal, financial) where labelled data is scarce. Combining domain-adaptive pre-traini
A Domain-adaptive Recurrent Neural Network (DA-RNN) is a recurrent neural network trained on a source domain and adapted to a target domain using domain adaptation techniques such as adversarial training, feature alignment, or fine-tuning. It enables sequential models to generalise across domains when labeled target-do
Domain-Adaptive Reinforcement Learning (DARL) extends standard RL by enabling a policy trained in one environment or domain to transfer and generalise effectively to a different but related target domain. It addresses the domain-shift problem — where dynamics, observations, or reward structures differ between training
Domain-adaptive RoBERTa-based classification extends the RoBERTa transformer by first continuing its masked-language-model pretraining on a domain-specific corpus before fine-tuning for a classification task. This two-stage adaptation bridges the gap between general web-crawled training data and specialized fields such
Domain-adaptive sentence embeddings extend general-purpose sentence encoders — such as Sentence-BERT — by continuing their training on domain-specific text. The result is a fixed-length vector representation that captures both universal language understanding and the vocabulary, style, and semantic nuances of the targe
Domain-adaptive sentiment analysis trains a sentiment model on one or more labeled source domains (e.g., product reviews) and adapts it to a target domain (e.g., social media posts or news) where labels are scarce or absent. By bridging the vocabulary and distributional gap between domains, it achieves strong sentiment
Domain-adaptive text summarization fine-tunes or adapts a pre-trained sequence-to-sequence language model on a target domain corpus so that summaries conform to domain-specific vocabulary, style, and factual constraints. It bridges the gap between general-purpose summarization models trained on news or web data and spe
A Domain-Adaptive Transformer (DAT) is a Transformer-based model — such as BERT or ViT — extended with an explicit domain-alignment objective so that learned representations transfer well from a labeled source domain to a different, often unlabeled, target domain. The approach combines the powerful representation capac
Domain-Adaptive Vision Transformer (DA-ViT) applies domain adaptation techniques — such as adversarial alignment, self-training, or attention-level bridging — on top of a pretrained Vision Transformer backbone to transfer visual knowledge from a labeled source domain to an unlabeled or lightly labeled target domain, re
Domain-adaptive Word2Vec trains or fine-tunes Word2Vec embeddings on a domain-specific text corpus so that word vectors capture the specialized vocabulary, semantic relationships, and jargon of a target field — such as clinical medicine, legal text, financial reports, or scientific literature — rather than reflecting g
Dosimetry measurement is the experimental quantification of radiation dose and exposure, originating from Röntgen and Becquerel's 1890s discoveries. It employs specialized detectors (ion chambers, TLD, Geiger counters) to measure photon and particle energy deposition in biological tissue or materials, providing direct
Drifter Lagrangian analysis tracks the motion of water parcels using surface drifters (buoys with attached drogues) to measure ocean currents directly. Developed by Robert Davis in the 1980s, this method provides direct observation of water parcel trajectories and enables estimation of eddy diffusivity, transport pathw
Droop Control is a decentralized control method that enables independent generators (inverters, microgrids) to operate synchronously without direct communication. Introduced by Guerrero et al. in 2013 for microgrids, droop control uses frequency and voltage deviations as signals to share power. By making generator outp
Dropout is a stochastic regularization technique for training deep neural networks, introduced by Srivastava, Hinton, Krizhevsky, Sutskever, and Salakhutdinov in 2014. During each training step, each neuron is independently switched off with probability (1 − p), preventing the network from co-adapting its units too tig
Dynamic Time Warping (DTW) is a sequence alignment algorithm that measures similarity between time series of different lengths by allowing flexible temporal matching. Applied to gait analysis, DTW enables comparison of walking patterns across subjects and conditions despite variations in cadence or stride length.
Dual-polarization (dual-pol) radar is a weather radar system that transmits and receives electromagnetic waves in both horizontal and vertical polarizations simultaneously. This technique, operational in weather services since the 2010s, provides detailed information about precipitation particle type, shape, and size d
The Dubins path is the shortest curve connecting two points in the plane with prescribed initial and terminal tangent directions, subject to a constraint on curvature. Introduced by Lester Dubins in 1957, it solved a fundamental problem in differential geometry and became essential in motion planning for aircraft, heli
The Dunn Index, introduced by Joseph C. Dunn in 1974, is a metric that captures cluster quality by measuring the ratio of the minimum between-cluster distance to the maximum within-cluster diameter. Higher values indicate well-separated and compact clusters, with better clustering quality.
Dynamic Application Security Testing (DAST) is a security analysis technique that tests a running application by sending various inputs and observing responses to identify vulnerabilities and security flaws. Developed in the 2000s as a complement to static analysis, DAST exercises the application at runtime, finding vu