Gabor Filters
A type of linear filter used in image processing and computer vision to analyze spatial frequency and orientation.
Types of Gabor Filters
- 2D Gabor Filters - Used for edge and texture detection.
- 3D Gabor Filters - Used in volumetric image analysis.
Example
Used in face recognition and fingerprint detection.
Gaussian Mixture Models (GMM)
A probabilistic model that represents data as a mixture of multiple Gaussian distributions.
Types of Gaussian Mixture Models
- Univariate GMM - Uses single-dimensional Gaussian distributions.
- Multivariate GMM - Uses multi-dimensional Gaussian distributions.
Example
Used in anomaly detection and speech recognition.
Gaussian Naïve Bayes
A variant of the Naïve Bayes classifier that assumes features follow a Gaussian distribution.
Types of Naïve Bayes Classifiers
- Gaussian Naïve Bayes - Assumes normally distributed features.
- Multinomial Naïve Bayes - Used for text classification.
Example
Used in spam detection and sentiment analysis.
Generative Adversarial Networks (GANs)
A class of machine learning models that use two neural networks (generator and discriminator) to generate realistic data.
Types of GANs
- Conditional GANs (cGANs) - Generates data based on specific conditions.
- CycleGAN - Converts images from one domain to another without paired examples.
Example
Used in image synthesis and deepfake generation.
Generalization in Machine Learning
The ability of a model to perform well on unseen data after training.
Types of Generalization
- Strong Generalization - Performs well on significantly different datasets.
- Weak Generalization - Performs well only on slightly varied datasets.
Example
Used in speech recognition systems to recognize different accents.
Genetic Algorithms
A type of evolutionary algorithm that uses natural selection principles to optimize machine learning models.
Types of Genetic Algorithms
- Simple Genetic Algorithm - Uses basic selection, crossover, and mutation.
- Parallel Genetic Algorithm - Uses multiple populations evolving simultaneously.
Example
Used in hyperparameter optimization and feature selection.
Gibbs Sampling
A Markov Chain Monte Carlo (MCMC) method for sampling from a complex probability distribution.
Types of Gibbs Sampling
- Collapsed Gibbs Sampling - Eliminates variables to improve efficiency.
- Blocked Gibbs Sampling - Samples multiple variables at once.
Example
Used in Bayesian inference and topic modeling.
Gini Impurity
A measure used in decision trees to determine the likelihood of incorrectly classifying a randomly chosen element.
Types of Decision Tree Impurity Measures
- Gini Impurity - Measures class distribution variance.
- Entropy - Measures information gain.
Example
Used in decision tree algorithms like CART.
Gradient Boosting
A boosting algorithm that improves weak learners iteratively by minimizing loss functions.
Types of Gradient Boosting
- XGBoost - Optimized for speed and performance.
- LightGBM - Uses histogram-based optimization.
Example
Used in financial risk modeling and fraud detection.
Graph Neural Networks (GNNs)
A class of deep learning models designed to process graph-structured data.
Types of Graph Neural Networks
- Graph Convolutional Networks (GCNs) - Uses spectral graph convolutions.
- Graph Attention Networks (GATs) - Uses attention mechanisms.
Example
Used in social network analysis and molecular property prediction.
Graph-Based Semi-Supervised Learning
A machine learning approach that uses graph structures to propagate labels from labeled to unlabeled data.
Types of Graph-Based Learning
- Label Propagation - Spreads labels across graph nodes.
- Manifold Regularization - Uses graph Laplacians to smooth predictions.
Example
Used in text classification and recommendation systems.
Graph Clustering
A method of partitioning a graph into clusters where nodes in the same group are more closely related.
Types of Graph Clustering
- Spectral Clustering - Uses eigenvalues of graph Laplacians.
- Community Detection - Identifies densely connected subgraphs.
Example
Used in social network analysis and fraud detection.
Graph Embeddings
A technique to represent graph nodes as numerical vectors while preserving structural properties.
Types of Graph Embeddings
- Node2Vec - Generates embeddings using random walks.
- GraphSAGE - Aggregates neighbor information for embeddings.
Example
Used in recommendation systems and fraud detection.
Graph Theory in Machine Learning
A mathematical framework for analyzing relationships between data points in graph structures.
Types of Graph Theory Applications
- Graph Neural Networks - Deep learning models for graphs.
- Graph-Based Clustering - Identifies hidden structures in data.
Example
Used in traffic network analysis and biology.
Greedy Algorithms
An optimization approach that makes the locally optimal choice at each step to achieve a global solution.
Types of Greedy Algorithms
- Prim's Algorithm - Finds minimum spanning trees.
- Huffman Coding - Used in data compression.
Example
Used in feature selection for machine learning models.
Grid Search
A hyperparameter optimization technique that exhaustively searches over predefined parameter sets.
Types of Hyperparameter Tuning
- Grid Search - Evaluates all parameter combinations.
- Random Search - Samples random combinations.
Example
Used in tuning hyperparameters for SVMs and deep learning models.
Group Lasso
A regularization technique that encourages sparsity at the group level instead of individual features.
Types of Lasso Regularization
- Standard Lasso - Shrinks individual coefficients.
- Group Lasso - Shrinks entire feature groups.
Example
Used in genomics to select relevant genes.
Growing Neural Gas (GNG)
An unsupervised neural network that dynamically adjusts its structure based on input data.
Types of Neural Gas Models
- Standard Neural Gas - Uses fixed network topology.
- Growing Neural Gas - Expands and adapts network structure.
Example
Used in clustering and dimensionality reduction.
Gaussian Process Regression (GPR)
A probabilistic regression method that models distributions over functions using Gaussian processes.
Types of Kernel Functions in GPR
- Radial Basis Function (RBF) Kernel - Models smooth functions.
- Matérn Kernel - Controls function roughness.
Example
Used in Bayesian optimization and time series forecasting.
Gradient Descent
An optimization algorithm used to minimize a function by iteratively moving in the direction of the steepest descent.
Types of Gradient Descent
- Batch Gradient Descent - Uses the entire dataset per update.
- Stochastic Gradient Descent - Updates weights per sample.
Example
Used in training deep learning models.
Gradient Clipping
A technique used to prevent exploding gradients by capping the gradients at a predefined threshold.
Types of Gradient Clipping
- Norm-Based Clipping - Limits the gradient norm.
- Value-Based Clipping - Caps gradients at a fixed value.
Example
Used in recurrent neural networks (RNNs) to stabilize training.
Gradient Checking
A technique to verify the correctness of computed gradients by comparing them with numerical approximations.
Types of Gradient Checking
- Finite Difference Method - Uses small perturbations to approximate gradients.
- Symbolic Differentiation - Uses algebraic methods to compute gradients.
Example
Used to debug backpropagation implementations in deep learning.
Gradient Noise
A regularization technique that adds noise to gradients to improve generalization.
Types of Gradient Noise
- Additive Noise - Adds random noise to gradients.
- Multiplicative Noise - Scales gradients with a noise factor.
Example
Used in deep learning to reduce overfitting.
Gradient Penalty
A regularization term that penalizes large gradients to stabilize training.
Types of Gradient Penalty
- Wasserstein GAN (WGAN-GP) - Uses gradient penalty for stability.
- Lipschitz Regularization - Constrains gradients to enforce smoothness.
Example
Used in training GANs for better convergence.
Gram Matrix
A matrix that captures similarity between feature vectors using dot products.
Types of Gram Matrices
- Kernel Gram Matrix - Used in kernel methods.
- Style Transfer Gram Matrix - Captures texture features.
Example
Used in neural style transfer and support vector machines.
Graph Attention Networks (GATs)
A type of graph neural network that uses attention mechanisms to weigh node relationships.
Types of Graph Neural Networks
- Graph Convolutional Networks (GCNs) - Uses spectral methods.
- Graph Attention Networks (GATs) - Uses attention mechanisms.
Example
Used in recommendation systems and molecular chemistry.
Graph Laplacian
A matrix representation of a graph used in spectral clustering and graph-based learning.
Types of Graph Laplacian
- Unnormalized Laplacian - Uses degree matrix and adjacency matrix.
- Normalized Laplacian - Scales eigenvalues for better stability.
Example
Used in spectral clustering and dimensionality reduction.
Graph Partitioning
A technique that divides a graph into smaller subgraphs while minimizing edge cuts.
Types of Graph Partitioning
- Edge Cut Partitioning - Minimizes the number of cut edges.
- Spectral Partitioning - Uses eigenvalues of the graph Laplacian.
Example
Used in parallel computing and network optimization.
Graph Signal Processing
An approach that extends signal processing concepts to graph-structured data.
Types of Graph Signal Processing
- Graph Fourier Transform - Generalizes classical Fourier analysis.
- Graph Wavelets - Provides localized spectral analysis.
Example
Used in sensor networks and biological data analysis.
Greedy Layer-Wise Pretraining
A training strategy that pretrains deep networks one layer at a time before fine-tuning.
Types of Layer-Wise Pretraining
- Autoencoder Pretraining - Uses unsupervised autoencoders.
- Restricted Boltzmann Machine (RBM) Pretraining - Uses energy-based models.
Example
Used in deep belief networks and transfer learning.
Greedy Search
A search algorithm that expands the most promising node at each step, aiming for an optimal solution.
Types of Greedy Search
- Best-First Search - Expands the node with the best heuristic value.
- A* Search - Combines greedy search with cost-based evaluation.
Example
Used in pathfinding algorithms like Dijkstra’s algorithm.
Grid-Based Clustering
A clustering method that divides data space into a finite number of grid cells and groups them based on density.
Types of Grid-Based Clustering
- STING (Statistical Information Grid) - Uses hierarchical grids.
- CLIQUE - Combines density-based and grid-based methods.
Example
Used in spatial data analysis and bioinformatics.
Growing Self-Organizing Maps
A variant of self-organizing maps (SOMs) that dynamically adjusts its structure based on data.
Types of Self-Organizing Maps
- Static SOM - Uses a fixed grid.
- Growing SOM - Adapts to data complexity.
Example
Used in feature extraction and dimensionality reduction.
Guided Backpropagation
A visualization technique that modifies standard backpropagation to highlight important input features.
Types of Backpropagation Techniques
- Standard Backpropagation - Computes gradients normally.
- Guided Backpropagation - Suppresses negative gradients.
Example
Used in deep learning to interpret convolutional neural networks.
Gaussian Mixture Model (GMM)
A probabilistic model that represents data as a mixture of multiple Gaussian distributions.
Types of GMMs
- Full Covariance GMM - Allows different covariance structures.
- Diagonal Covariance GMM - Assumes independent features.
Example
Used in anomaly detection and image segmentation.
Generative Adversarial Networks (GANs)
A deep learning model consisting of a generator and a discriminator that compete to generate realistic data.
Types of GANs
- DCGAN - Uses convolutional layers for better image generation.
- WGAN - Uses Wasserstein loss for more stable training.
Example
Used in deepfake generation and image super-resolution.
Genetic Algorithms
An optimization technique inspired by natural selection that evolves solutions through mutation and crossover.
Types of Genetic Algorithms
- Simple Genetic Algorithm (SGA) - Uses standard selection and mutation.
- Adaptive Genetic Algorithm (AGA) - Adjusts mutation and crossover rates dynamically.
Example
Used in feature selection and neural architecture search.
Graph Neural Networks (GNNs)
A deep learning framework designed to process graph-structured data.
Types of GNNs
- Graph Convolutional Networks (GCNs) - Uses spectral convolution.
- Graph Attention Networks (GATs) - Applies attention mechanisms to graphs.
Example
Used in social network analysis and recommendation systems.
Granger Causality
A statistical method to determine whether one time series can predict another.
Types of Granger Causality
- Linear Granger Causality - Uses linear regression models.
- Nonlinear Granger Causality - Uses machine learning-based models.
Example
Used in econometrics and time-series forecasting.
Gaussian Process Latent Variable Model (GP-LVM)
A probabilistic dimensionality reduction technique that models data using Gaussian processes.
Types of GP-LVM
- Standard GP-LVM - Models latent variables using Gaussian processes.
- Bayesian GP-LVM - Uses Bayesian inference for uncertainty estimation.
Example
Used in visualization of high-dimensional data.
Machine Learning (ML)
ML is a subset of AI that enables machines to learn patterns from data and make predictions or decisions without explicit programming.
Types of ML
- Supervised Learning
- Unsupervised Learning
- Reinforcement Learning
Example
Spam detection in emails using classification models.
Deep Learning (DL)
DL is a subset of ML that uses artificial neural networks to process complex data and perform high-level computations.
Example
Image recognition in self-driving cars.
Generative AI (Gen AI)
Gen AI refers to AI models that generate new content, including text, images, and code, using trained knowledge bases.
Example
AI models like ChatGPT and Stable Diffusion that generate text and images.