按方法、领域与证据探索科学。

Federated Learning

Federated Learning is a distributed machine learning paradigm introduced by McMahan et al. in 2017 in which a global model is trained collaboratively across multiple decentralized clients — such as mobile devices or hospital systems — without ever transferring raw data to a central server. Each participant computes mod

deep learning

FEDformer

FEDformer is a Transformer-based architecture for long-term multivariate time-series forecasting, introduced by Zhou et al. at ICML 2022. Its core innovation is the combination of seasonal-trend decomposition with frequency-domain attention: instead of computing full token-to-token attention in the time domain, FEDform

control theory

Feedback Linearization

Feedback Linearization is a nonlinear control technique that uses a nonlinear state-feedback transformation to convert a nonlinear system into a linear one, enabling the use of standard linear control methods. Developed by Isidori, Sontag, and others in the 1980s, feedback linearization is conceptually elegant and powe

machine learning

Few-shot Learning

Few-shot learning is a machine learning paradigm that trains models to recognize new classes or solve new tasks from only a handful of labeled examples — typically one to five — by leveraging prior knowledge acquired from a large, related training distribution. It is especially relevant in domains where labeling is exp

deep learning

Few-Shot Object Detection

Few-Shot Object Detection (FSOD) is a meta-learning approach that enables detecting novel object classes from only a few annotated examples. Unlike standard object detection requiring hundreds of labeled instances per class, FSOD learns to quickly adapt detection models to new object categories by leveraging knowledge

text mining

Few-Shot Text Classification

Few-shot text classification assigns documents to classes using only a handful of labelled examples per class. Building on advances by Gao et al. (2021) and the prompt-free SetFit approach of Tunstall et al. (2022), it leans on prototypical networks, MAML, or fine-tuning of a large pretrained model to learn from scarce

particle physics

Feynman Diagram

Feynman diagrams are graphical representations of particle interactions introduced by Richard Feynman in 1949. They provide an intuitive and systematic way to visualize and calculate amplitudes for quantum field theory processes, converting complex mathematical expressions into geometric pictures that reveal the underl

thermodynamics

Fick's Laws

Fick's Laws describe how species diffuse through media due to concentration gradients. The First Law (steady-state) relates diffusion flux to concentration gradient, while the Second Law (transient) describes how concentration changes over time. These laws are fundamental to mass transfer analysis, applying to gases, l

control theory

Field-Oriented Control

Field-Oriented Control (FOC), also known as Vector Control, is an advanced method for controlling AC induction and permanent magnet motors by decomposing phase currents into torque and flux components and independently regulating them using PI controllers. Pioneered by Blaschke in 1972, FOC enables smooth precise motor

deep learning

FiLM

FiLM is a long-term time-series forecasting architecture introduced by Tian Zhou and colleagues at NeurIPS 2022. It combines Legendre polynomial projections of the historical input with learnable frequency-domain filters applied to the resulting coefficient sequences. By representing history as a compact set of polynom

deep learning

Fine-Tuned BERT-based Classification

Fine-Tuned BERT-based Classification adapts a pre-trained BERT transformer to a specific text classification task by adding a lightweight output layer and continuing gradient-based training on labelled examples. It consistently achieves near-state-of-the-art accuracy on sentiment analysis, topic categorisation, intent

deep learning

Fine-Tuned Convolutional Neural Network

Fine-tuning a CNN means starting from a network already trained on a large dataset — typically ImageNet — and continuing training on a smaller target dataset so the model adapts its learned visual features to a new task. This approach dramatically reduces the data and compute required to reach strong performance compar

deep learning

Fine-Tuned Diffusion Model

A fine-tuned diffusion model adapts a large pretrained denoising diffusion model — such as Stable Diffusion or DALL-E — to a specific subject, style, or domain by continuing training on a small curated dataset. Techniques such as DreamBooth, textual inversion, and LoRA make this adaptation feasible on consumer hardware

deep learning

Fine-Tuned Doc2Vec

Fine-Tuned Doc2Vec adapts a pre-trained Paragraph Vector (Doc2Vec) model by continuing its training on a target corpus, producing document embeddings that capture both the general language knowledge of the original training and the vocabulary and style of the new domain. It is used for text classification, semantic sim

deep learning

Fine-Tuned Generative Adversarial Network

A Fine-Tuned GAN starts from a large pre-trained generative adversarial network and continues adversarial training on a smaller target dataset, allowing the model to synthesize high-quality samples in a new domain without training from scratch. This transfer approach dramatically reduces data and compute requirements w

deep learning

Fine-Tuned GRU

Fine-Tuned GRU adapts a Gated Recurrent Unit network — pre-trained on a large source dataset — to a specific target task or domain by continuing training on domain-specific labeled data. This combines the sequential memory capacity of GRUs with the efficiency gains of transfer learning, achieving strong performance eve

deep learning

Fine-Tuned Image Classification

Fine-tuned image classification adapts a large neural network pretrained on a broad image corpus (such as ImageNet) to a specific target domain by continuing training on labeled domain images. This approach achieves strong accuracy with far fewer target-domain samples than training from scratch, making it the dominant

deep learning

Fine-Tuned LDA Topic Model

Fine-Tuned LDA adapts a Latent Dirichlet Allocation model trained on a large general corpus to a specific target domain by continuing inference on domain-specific documents. Rather than fitting LDA from scratch, the pre-trained topic-word distributions are used as an informed starting point, enabling the model to disco

deep learning

Fine-Tuned LSTM

Fine-Tuned LSTM adapts a Long Short-Term Memory network pre-trained on a large corpus to a specific downstream task — such as text classification, sentiment analysis, or sequence labeling — by continuing training on task-specific labeled data. Popularised by the ULMFiT framework, this approach achieves strong performan

deep learning

Fine-Tuned Multilayer Perceptron

A Fine-Tuned Multilayer Perceptron starts from weights learned on a source task — or a large general-purpose dataset — and continues training on a smaller target dataset with a reduced learning rate. This reuse of pre-learned representations allows the MLP to converge faster and generalise better than training from scr

deep learning

Fine-Tuned Named Entity Recognition

Fine-Tuned Named Entity Recognition adapts a pre-trained language model — most commonly BERT or one of its derivatives — to the task of identifying and classifying named entities (persons, organizations, locations, dates, etc.) in text. By fine-tuning on a relatively small labeled corpus, practitioners achieve state-of

deep learning

Fine-Tuned Question Answering

Fine-Tuned Question Answering adapts a large pre-trained language model — such as BERT, RoBERTa, or a GPT-family model — to answer natural-language questions over a given context passage or knowledge base. The model learns to locate answer spans or generate free-form answers by continuing training on labeled QA pairs a

deep learning

Fine-Tuned Recurrent Neural Network

A Fine-Tuned Recurrent Neural Network (RNN) starts from a model pre-trained on large corpora or time-series data and adapts its weights to a specific downstream task through controlled gradient updates. The approach dramatically cuts the labeled data needed for strong sequence modeling performance in text classificatio

deep learning

Fine-Tuned Reinforcement Learning

Fine-Tuned Reinforcement Learning adapts a pre-trained policy or model to a new task or behavioral objective using reinforcement signals — including human feedback — rather than retraining from scratch. Popularized by RLHF, it is the core technique behind aligning large language models and adapting deep RL agents to sp

deep learning

Fine-Tuned RoBERTa-based Classification

Fine-tuned RoBERTa-based classification adapts the RoBERTa pretrained transformer — itself a robustly retrained variant of BERT — to a specific text classification task by appending a classification head and continuing training on labeled examples. It consistently achieves state-of-the-art or near-state-of-the-art perf

deep learning

Fine-Tuned Semantic Segmentation

Fine-tuned semantic segmentation adapts a deep neural network pre-trained on a large pixel-labelled dataset (e.g., ImageNet-pretrained backbone with an encoder-decoder head trained on COCO or Cityscapes) to a new target domain by continuing training on domain-specific annotated images. The result is a model that assign

deep learning

Fine-Tuned Sentence Embeddings

Fine-Tuned Sentence Embeddings adapt a general-purpose pre-trained sentence encoder — such as Sentence-BERT — to a specific domain or task by continuing training on labeled or paired text data from that domain. The resulting embeddings capture domain-specific semantic structure far better than off-the-shelf vectors, im

deep learning

Fine-Tuned Text Summarization

Fine-Tuned Text Summarization adapts a large pre-trained sequence-to-sequence model — such as BART, T5, or PEGASUS — to generate concise summaries of documents by training on domain-specific (document, summary) pairs. The approach yields substantially more fluent and faithful summaries than extractive or generic approa

deep learning

Fine-Tuned Topic Modeling

Fine-Tuned Topic Modeling adapts pre-trained language models — such as BERT or Sentence-BERT — to discover latent topics in document collections. Unlike classical probabilistic methods (LDA, NMF), it leverages rich contextual embeddings and optionally fine-tunes the backbone on domain-specific corpora, producing more c

deep learning

Fine-Tuned Transformer

Fine-tuning a Transformer adapts a large pre-trained model — such as BERT, GPT, or ViT — to a specific downstream task by continuing gradient-based training on a labelled target dataset. This two-stage paradigm (pre-train then fine-tune) consistently achieves state-of-the-art results across NLP and computer vision task

deep learning

Fine-Tuned Variational Autoencoder

A Fine-Tuned Variational Autoencoder begins with a VAE pre-trained on a large source dataset and then continues training on a smaller target-domain dataset. This approach adapts the learned latent representation and generative capacity to new data, preserving general structure while specializing to the target distribut

deep learning

Fine-Tuned Vision Transformer

Fine-Tuned Vision Transformer adapts a large pre-trained ViT model — which splits images into fixed-size patches and processes them through self-attention layers — to a new image classification or recognition task using a relatively small labeled dataset. It achieves state-of-the-art accuracy in computer vision by leve

deep learning

Fine-Tuned Word2Vec

Fine-Tuned Word2Vec adapts a pre-trained Word2Vec model to a specific domain or task by continuing its training on domain-specific text. Rather than training embeddings from scratch, practitioners load general-purpose vectors (e.g., Google News embeddings) and run additional Skip-gram or CBOW epochs on domain corpora,

reliability engineering

Finite Element Model Updating

Finite Element Model (FEM) Updating is the process of refining a numerical structural model to match measured behavior (modal properties, vibrations, static displacements) from the physical structure. By comparing computational predictions to experimental data and systematically adjusting uncertain model parameters (ma

electrical engineering

Finite Integration Technique

The Finite Integration Technique (FIT) is a numerical method for solving Maxwell equations on structured grids, formulating electromagnetics as a system of integral equations over grid cells. Introduced by Thomas Weiland in 1977, FIT bridges finite differences and finite elements, offering excellent accuracy, stability

civil engineering

Finite Strip Method

The finite strip method (FSM) is a semi-analytical numerical approach for analyzing prismatic or cylindrical structures by dividing them into strips in one direction and using analytical or exact solutions in the perpendicular direction. Developed by Cheung in 1976, FSM reduces computational cost and often provides sup

thermodynamics

Finite-Time Thermodynamics

Finite-Time Thermodynamics (FTT) relaxes the classical assumption that thermodynamic processes occur reversibly (infinitely slowly). Instead, it analyzes real thermal systems operating at finite rates with irreversibilities. FTT reveals fundamental trade-offs: to complete a process quickly requires accepting large irre

mining engineering

Flotation Kinetics

Flotation kinetics is the study of how recovery of minerals from ore changes over time during flotation. The Garcia-Zuniga model, introduced in 1935, describes recovery as a first-order kinetic process with rate constant k and maximum recoverable fraction R∞. This simple model underpins flotation cell design and proces

biomechanics

Forward Kinematics

Forward kinematics is the calculation of the position and orientation of a distal body segment (such as the hand) based on the joint angles of proximal segments. Originally formalized in robotics by John Craig and adapted to biomechanics, it allows practitioners to predict endpoint location from known joint configurati

optics

Fourier Optics

Fourier optics is a mathematical framework that analyzes optical systems and phenomena using Fourier transforms and frequency-domain methods. Grounded in Joseph Fourier's 1822 work on heat diffusion and Ernst Abbe's microscopy theory, this approach decomposes optical fields into plane waves or spatial frequencies, reve

signal processing

Fourier Transform

The Fourier Transform decomposes a time-domain signal into its constituent sinusoidal frequencies, revealing the spectral content hidden within complex waveforms. Joseph Fourier introduced the continuous transform in 1822, but the computationally efficient Fast Fourier Transform (FFT) was formalized by James Cooley and

model evaluation

Fowlkes-Mallows Index

The Fowlkes-Mallows Index, introduced by Fowlkes and Mallows in 1983, is an external clustering evaluation metric based on the geometric mean of precision and recall. It measures agreement between two partitions by examining pairs of points and how they are grouped in both the predicted and ground truth clusterings. Va

machine learning

FP-Growth

FP-Growth, introduced by Jiawei Han, Jian Pei, and Yiwen Yin in 2000, mines frequent itemsets from transaction data without generating candidate sets, the costly step that slows the classic Apriori algorithm. It compresses the database into a frequent-pattern tree (FP-tree) in two scans, then grows frequent patterns re

complex systems

Fractal Analysis

Fractal Analysis quantifies the self-similar, scale-invariant complexity of geometric objects and time series through the fractal dimension D and the Hurst exponent H. Introduced systematically by Benoit Mandelbrot in his 1983 landmark work, the framework extends classical Euclidean geometry to irregular shapes found i

text mining

Frame Analysis

Frame analysis is a FrameNet-based natural-language-processing task that detects the semantic frames evoked in text and the participant roles (frame-evoking elements and frame elements, FE) that fill them. Rooted in Charles Fillmore's frame semantics (1982) and operationalised by the Berkeley FrameNet Project (Baker et

deep learning

FreTS

FreTS is a time series forecasting architecture introduced by Yi et al. at NeurIPS 2023. It departs from Transformer-based designs by applying simple Multi-Layer Perceptrons (MLPs) entirely in the frequency domain. The model transforms input sequences with the Discrete Fourier Transform and then learns temporal and cha

spectroscopy

FT-ICR Mass Spectrometry

Fourier Transform Ion Cyclotron Resonance (FT-ICR) mass spectrometry is an advanced analytical technique that combines magnetic confinement of ions with Fourier transform data processing to achieve exceptional mass accuracy and resolution. Developed by Comisarow and Marshall in 1974, FT-ICR-MS enables the determination

deep learning

Fully Convolutional Network (FCN)

The Fully Convolutional Network (FCN), introduced by Long, Shelhamer, and Darrell at CVPR 2015, was the first end-to-end deep learning architecture trained to produce dense pixel-wise semantic segmentation maps from images of arbitrary size. By replacing the fully connected layers of a classification CNN with convoluti

software engineering

Function Point Analysis

Function point analysis (FPA) quantifies software size by counting business functions and user interactions independent of technology or programming language. Introduced by Albrecht (1979), FPA measures delivered functionality, enabling effort estimation, productivity benchmarking, and software value assessment. Organi

cryptography

Fuzzing

Fuzzing is a software testing technique that inputs large numbers of random or semi-random test cases to a program to find bugs, crashes, and security vulnerabilities. Pioneered by Barton Miller in 1990, fuzzing has become a primary method for discovering zero-day vulnerabilities in complex software. Modern fuzzing too

machine learning

Fuzzy C-Means

Fuzzy C-Means is a soft clustering algorithm in which every data point belongs to every cluster with a graded membership between 0 and 1, rather than being assigned to exactly one cluster. Originated by Joseph Dunn in 1973 and generalized by James Bezdek in 1981, it minimizes a fuzzy-weighted within-cluster variance, m

acoustics

FxLMS Active Noise Control

The Filtered-x Least Mean Squares (FxLMS) algorithm is an adaptive filter used in active noise control (ANC) systems to reduce unwanted sound by generating anti-noise. Pioneered by Widrow and Stearns in 1975 and refined by Eriksson and colleagues, FxLMS is the most widely deployed algorithm in commercial noise-cancelin

numerical methods

Galerkin Method

The Galerkin Method is a projection-based variational technique for solving differential equations by reducing infinite-dimensional problems to finite-dimensional linear systems. Developed by Boris Galerkin in 1915 and independently by Ivan Bubnoff, it underpins the Finite Element Method (FEM) and is foundational to mo

model evaluation

Gap Statistic

The Gap Statistic, developed by Tibshirani, Walther, and Hastie in 2001, is a principled statistical method for determining the optimal number of clusters in a dataset. It compares the observed within-cluster sum of squares to the expected value under a null hypothesis of no clustering structure, providing a theoretica

deep learning

Gated Recurrent Unit

The Gated Recurrent Unit (GRU), introduced by Cho et al. in 2014, is a streamlined recurrent neural network that uses two learned gates — an update gate and a reset gate — to selectively retain or discard information across time steps, enabling effective sequence modelling with fewer parameters than LSTM.

machine learning

Gaussian Mixture Model

A Gaussian Mixture Model is a probabilistic clustering method that models the data as a weighted mixture of several Gaussian distributions, fitted with the Expectation–Maximization algorithm formalized by Dempster, Laird & Rubin in 1977. It is a generalization of K-means in which each cluster can take its own shape, si

genetics

GCTA

GCTA (Genome-wide Complex Trait Analysis) is a computational toolkit for estimating heritability and genetic correlations from genome-wide genotype and phenotype data. Developed by Yang and Visscher in 2011, GCTA uses genome-wide restricted maximum likelihood (GREML) to partition phenotypic variance into components exp

particle physics

Geant4 Simulation

Geant4 is a Monte Carlo simulation toolkit for the passage of particles through matter, developed by an international collaboration. It provides a comprehensive framework for modeling detector geometries, simulating particle interactions, and predicting detector responses, making it essential for designing and optimizi

text mining

Gender Bias Detection

Gender bias detection in NLP is a family of statistical and embedding-based methods used to measure stereotyping, representational imbalance, and occupational bias in text corpora and language models. Grounded in benchmarks established by Caliskan et al. (2017) with the Word Embedding Association Test (WEAT) and Zhao e

bioinformatics

Gene Set Enrichment Analysis

Gene Set Enrichment Analysis (GSEA) is a computational method that determines whether a predefined set of genes — representing a biological pathway, process, or function — shows statistically significant, coordinated differences between two biological conditions. Unlike simple fold-change filtering, GSEA operates on al