แคตตาล็อกวิธีวิจัยเพียงหนึ่งเดียว — เรียนรู้ว่าแต่ละวิธีทำงานอย่างไร ใช้เมื่อใด และทำอะไรไม่ได้
Multimodal LSTM extends the standard Long Short-Term Memory network to jointly process sequential data from multiple input modalities — such as text, audio, and video — within a unified recurrent architecture. By fusing representations from different sources before or within the LSTM cells, it captures temporal depende
A Multimodal Multilayer Perceptron (MM-MLP) is a feedforward neural network that ingests features from two or more heterogeneous input modalities — such as structured tabular data, text embeddings, and image feature vectors — by encoding each stream separately and fusing them into a shared representation before passing
Multimodal Named Entity Recognition (MNER) extends classical NER by fusing textual sequences with complementary modalities — most commonly images — to improve the identification and classification of named entities such as persons, organizations, and locations in settings where visual context disambiguates ambiguous or
Multimodal NLP is a family of natural-language-processing pipelines that combine text with one or more additional data modalities — most commonly images, but also audio and video — to perform understanding and generation tasks such as visual question answering, image captioning, and multimodal sentiment recognition. Th
Multimodal NMF Topic Model extends Non-negative Matrix Factorization to simultaneously discover latent topics across multiple data modalities — such as text and images — by enforcing shared or aligned low-rank factor matrices. It uncovers coherent, interpretable topics that jointly explain patterns in both textual and
Multimodal object detection extends single-modality object detectors by jointly processing signals from multiple sensor types — such as RGB cameras, depth sensors, LiDAR, radar, or text descriptions — to localize and classify objects with higher accuracy and robustness than any single modality alone. Fusion of compleme
Multimodal question answering (Multimodal QA) is a class of deep-learning methods that answer natural-language questions by jointly reasoning over information from multiple modalities — most commonly text and images, but also video, audio, and structured tables. Introduced prominently through the VQA benchmark in 2015,
A Multimodal Recurrent Neural Network combines inputs from two or more data modalities — such as images, text, and audio — within a recurrent sequence-processing framework. It encodes each modality separately, fuses the representations, and then processes the combined signal through recurrent units (RNN, LSTM, or GRU)
Multimodal Reinforcement Learning trains agents to make sequential decisions by perceiving and integrating multiple input modalities — such as raw pixels, language instructions, audio, and proprioceptive sensors — simultaneously. Rather than acting on a single data stream, the agent fuses heterogeneous signals into a u
Multimodal RoBERTa-based Classification combines the RoBERTa transformer encoder — a robustly optimised variant of BERT — with auxiliary modalities such as images, structured metadata, or tabular features. The fused representation is passed to a classification head, allowing the model to leverage both rich language und
Multimodal semantic segmentation assigns a semantic class label to every pixel in a scene by fusing information from two or more sensor modalities — most commonly RGB images paired with depth maps (RGB-D), LiDAR point clouds, thermal cameras, or text descriptions. Deep encoder-decoder networks learn to align and fuse c
Multimodal sentence embeddings map text and images (and sometimes audio or video) into a shared continuous vector space, so that semantically related pairs from different modalities land close together. Trained by contrastive objectives on large paired corpora, these representations power cross-modal retrieval, zero-sh
Multimodal text summarization generates a concise textual summary by jointly processing multiple input modalities — most commonly text and images, but also video frames or audio — using deep learning models that align visual and linguistic representations. The output is a natural-language summary that captures salient
Multimodal topic modeling discovers latent thematic structure shared across multiple data modalities — for example, co-occurring words and images — by learning a joint probabilistic representation that aligns topics across modalities. It extends classical text-only approaches such as LDA to settings where each document
A Multimodal Transformer extends the standard Transformer architecture to process and jointly reason over two or more input modalities — most commonly text and images, but also audio, video, or structured data. Cross-modal attention layers allow information from one modality to inform representations in another, enabli
The Multimodal Variational Autoencoder (MVAE) is a deep generative model that learns a shared latent representation across two or more data modalities — such as images and captions — using a product-of-experts fusion of modality-specific encoders, enabling generation and inference even when only a subset of modalities
Multimodal Vision Transformer (Multimodal ViT) extends the Vision Transformer architecture to jointly process and align representations from multiple modalities — typically images and text — using self-attention and cross-attention mechanisms. By learning shared or aligned embedding spaces across modalities, it enables
Multimodal Word2Vec extends the classic Word2Vec framework by grounding word representations in perceptual signals — typically image features — alongside distributional text statistics. The result is word vectors that capture both linguistic co-occurrence patterns and visual meaning, enabling richer semantic similarity
Multiplex network analysis studies systems where the same set of nodes is connected by multiple distinct types of relationships, each represented as a separate network layer. By analyzing layers simultaneously rather than in isolation, it reveals how different relation types interact, reinforce each other, or compensat
Multitask Learning (MTL) is a machine learning paradigm in which a model is trained simultaneously on multiple related tasks, sharing representations across them to improve generalization. Introduced formally by Rich Caruana in 1997, MTL draws on the intuition that auxiliary tasks act as inductive bias, providing extra
Muscle synergy analysis decomposes complex motor behavior into a small set of coactivated muscle groups (synergies or motor primitives). Pioneered by Marc Tresch and colleagues studying frog motor control, this approach reveals how the nervous system simplifies the control of many muscles by organizing them into task-r
Music genre classification is the task of automatically assigning genre labels (rock, jazz, classical, pop, etc.) to audio recordings. Introduced formally by Tzanetakis and Cook (2002), it is one of the earliest and most studied music information retrieval problems. It remains critical for music discovery, recommendati
Music segmentation is the task of dividing a musical recording into distinct structural sections (e.g., verse, chorus, bridge, pre-chorus, outro). Introduced by Goto (2001), it identifies major structural boundaries and labels sections according to musical form. Segmentation is essential for music understanding, audio
Music similarity measures are computational methods for assessing how musically related two audio recordings are. Introduced by Logan (2001), similarity measures enable content-based music recommendation, playlist generation, and music discovery. Unlike fingerprinting, which identifies the same song, similarity measure
Musical key detection is the task of automatically determining the key (tonal center) and scale mode of a musical composition from its audio. Introduced formally by Gómez (2006), it is essential for music analysis, transposition, harmonic understanding, and music theory education. The key defines the tonal center aroun
The Muskingum method is a hydrologic flood routing technique that predicts how a flood wave attenuates (reduces in peak) and spreads as it travels down a river reach. Developed by McCarthy in 1938 for the US Army Corps of Engineers, the method is simple enough for hand calculations while capturing the essential physics
N-BEATS is a deep learning architecture for time series forecasting, introduced by Oreshkin and colleagues in 2020, built from interpretable trend and seasonality stacks. It was the first purely neural forecasting model to reach state-of-the-art performance on the M4 competition without relying on any classical statist
N-BEATSx is an extension of the N-BEATS neural time series forecasting model that incorporates exogenous (external) variables through a cross-learner architecture. Published in 2023, N-BEATSx improves upon N-BEATS by enabling the model to leverage additional features beyond the historical time series values.
N-body simulation is a computational method for modeling the dynamics of a system of particles under mutual gravitational forces. Originating from Newton's laws of motion and gravitation, it solves the fundamental equations of celestial mechanics. This technique is essential for understanding planetary orbits, star clu
An n-gram language model is a statistical model that predicts the probability of the next word by looking only at the previous n−1 words. Described in detail by Jurafsky and Martin (Speech and Language Processing), it provides foundational infrastructure for text generation, spelling correction, and speech recognition.
N-HiTS (Neural Hierarchical Interpolation for Time Series Forecasting), introduced by Challu and colleagues in 2023, is a deep neural forecasting architecture that combines the hierarchical forecasts of multiple stacks operating at different sampling rates and merges them through interpolation. It extends N-BEATS to de
Naive Bayes is a fast probabilistic classifier that applies Bayes' theorem while assuming that the features are conditionally independent given the class — a method given its standard machine-learning treatment in Tom Mitchell's 1997 textbook Machine Learning. Despite this simplifying ('naive') assumption, it is quick
Named entity recognition (NER) is a natural-language-processing task that automatically detects and labels entities in text — such as people, organisations, locations, and dates. Surveyed by Nadeau and Sekine (2007) and later advanced with neural architectures by Lample et al. (2016), it turns free-running text into ta
Nanoindentation, or instrumented indentation, is a technique for measuring the hardness and elastic modulus of materials by pressing a hard probe into a sample surface and continuously recording load and penetration depth. Developed by Oliver and Pharr in 1992, nanoindentation enables measurement of mechanical properti
Natural Language Generation (NLG) is the branch of natural language processing that automatically produces fluent, human-readable text from structured data, knowledge graphs, or semantic representations. Formalised in the classical pipeline by Reiter and Dale (2000) and surveyed comprehensively by Gatt and Krahmer (201
The Normalized Difference Vegetation Index (NDVI) is a spectral index computed from satellite or aerial multispectral imagery that quantifies vegetation greenness and vigor. Introduced by Rouse and colleagues in 1973 using Landsat data, NDVI has become the most widely used remote sensing metric for vegetation monitorin
NEAT is a genetic algorithm for evolving artificial neural networks introduced by Kenneth Stanley and Risto Miikkulainen in 2002. Unlike methods that evolve weights alone, NEAT simultaneously evolves both the topology (structure) and the connection weights of neural networks. It achieves this through a direct genome en
Negation detection is a natural-language-processing task that locates negation cues in text — words or phrases such as 'no', 'not', 'without', or 'denies' — and determines the span of text (the scope) whose meaning those cues invert. Formalised for clinical text by Chapman et al. (2001) with the NegEx algorithm and ext
Network diffusion analysis models how information, diseases, behaviors, or innovations spread across a graph of nodes and edges. Drawing on classical epidemic theory (SI, SIR, SIS) and modern network science, it tracks which nodes become infected, how quickly, and whether the spread reaches a global cascade or dies out
Network embedding is a family of representation-learning methods that map each node of a graph into a dense, low-dimensional vector while preserving the network's structural properties. The approach was formalised for social-network data by Perozzi, Al-Rfou, and Skiena with DeepWalk (2014), which adapted the Word2Vec s
Network motif analysis is a statistical method for directed networks, introduced by Milo, Shen-Orr, and Alon in 2002, that identifies small recurring subgraph patterns — motifs — that appear significantly more often than would be expected in a comparable random network. By comparing a real network against a null ensemb
Network resilience and vulnerability analysis is an analytical framework, formalised by Albert, Jeong, and Barabási (2000), that measures how a network degrades functionally as nodes or edges are progressively removed. By running targeted-attack simulations — removing the highest-centrality nodes first — and random-fai
Network-based copy number variation analysis integrates genome-wide CNV data with biological interaction networks — such as protein-protein interaction (PPI) or pathway networks — to identify functionally coherent regions, driver genes, and altered subnetworks that raw CNV calling alone would miss. By propagating CNV s
Network-based EWAS extends conventional epigenome-wide association studies by overlaying differentially methylated positions or regions onto biological interaction networks — such as protein-protein interaction, co-expression, or gene regulatory networks — to identify functionally coherent epigenetic modules rather tha
Network-based eQTL analysis extends classical eQTL mapping by embedding genetic variant-to-expression associations within gene regulatory or protein interaction networks. Rather than treating each SNP-gene pair independently, this approach leverages network topology — such as co-expression modules or known pathway stru
Network-based GWAS integrates conventional genome-wide association study results with biological network data — such as protein-protein interaction (PPI) networks or gene co-expression graphs — to identify disease-relevant gene modules or subnetworks. Instead of reporting only the top individual SNPs, this approach pro
Network-based metabolomics analysis integrates quantitative metabolite profiling data with biological network structures — metabolic pathways, protein-metabolite interaction graphs, and disease networks — to reveal coordinated biochemical disruptions that individual metabolite lists would miss. Rather than treating eac
Network-based microbiome diversity analysis integrates graph-theoretic co-occurrence network inference with classical alpha- and beta-diversity metrics to characterize the structural organization of microbial communities. Rather than treating taxa as independent entities, the method models pairwise microbial associatio
Network-based pathway enrichment analysis integrates molecular interaction networks — protein-protein interactions, signalling graphs, or gene regulatory networks — with omics measurements to identify biological pathways that are coordinately altered in a condition. Unlike classical over-representation or gene-set enri
Network-based phylogenetic analysis constructs graph-structured representations of evolutionary relationships that explicitly accommodate reticulate events — including hybridization, horizontal gene transfer, recombination, and incomplete lineage sorting — which strictly bifurcating phylogenetic trees cannot represent.
Network-based RNA-seq differential expression analysis integrates conventional differential expression testing with gene interaction networks — such as protein-protein interaction graphs or weighted co-expression networks — to identify not just individual differentially expressed genes but coherent, biologically meanin
Network-based single-cell RNA-seq analysis extends standard scRNA-seq workflows by constructing and interrogating molecular interaction networks — gene regulatory networks, co-expression networks, or cell-cell communication graphs — from single-cell transcriptomic data. Rather than treating each gene independently, thi
Network-based (graph-genome) variant calling replaces the conventional single linear reference genome with a variation graph — a network in which nodes represent sequence segments and edges represent known alternative paths through the genome. Reads are mapped onto this graph, enabling detection of SNPs, indels, and st
Neural Architecture Search (NAS), introduced by Zoph and Le in 2017, automatically optimizes architectural decisions such as a network's depth, width, and connection structure instead of hand-designing them. Leading methods in the field include DARTS, ENAS, and Once-for-All.
A Neural ODE, introduced by Chen and colleagues in 2018, models a hidden state as the continuous solution of an ordinary differential equation whose dynamics are parameterised by a neural network. It generalises the limiting case of residual connections, making it well suited to irregularly spaced time series and physi
Neural Radiance Fields (NeRF) is a method introduced by Mildenhall et al. in 2020 that represents a 3D scene as a continuous function parameterized by a neural network. Given multi-view images of a scene, NeRF learns to predict the color and density of light rays at any spatial location and viewing angle, enabling nove
Neural Style Transfer (NST) is a deep-learning image synthesis technique, introduced by Gatys, Ecker, and Bethge in 2015, that separates the semantic content of one image from the visual texture and artistic style of another, then recombines them into a single synthesized image by iteratively optimizing pixel values to
Neutrino oscillation analysis is the study of flavor mixing in the neutrino sector, where neutrinos born as one flavor (electron, muon, or tau) spontaneously convert into other flavors as they propagate. Measuring oscillation parameters provides crucial evidence for physics beyond the Standard Model and tests our under
Neutron activation analysis (NAA) is an analytical technique for determining elemental composition by bombarding samples with neutrons to produce radioactive isotopes, invented by de Hevesy and Levi in 1936. By measuring decay gamma rays from irradiated samples, NAA quantifies trace and major elements with high sensiti
Neutron transport calculation is a computational method for determining the distribution and behavior of neutrons in a nuclear medium, developed during the Manhattan Project in the 1940s. It solves the Boltzmann transport equation to predict neutron flux, energy spectra, and reaction rates essential for reactor design