Haar Wavelet
A mathematical function used in wavelet transforms for feature extraction and signal compression.
Types of Haar Wavelet Transforms
- Discrete Haar Wavelet Transform (DHWT) - Used for image compression.
- Continuous Haar Wavelet Transform - Used in signal processing.
Example
Used in JPEG2000 image compression.
Hard Attention
An attention mechanism that focuses on specific parts of the input while ignoring others.
Types of Attention Mechanisms
- Soft Attention - Weighs all parts of the input.
- Hard Attention - Selects a subset of input regions.
Example
Used in object detection models like YOLO.
Hamming Distance
A metric that measures the number of different bits between two binary strings.
Types of Distance Metrics
- Hamming Distance - Counts differing bits.
- Euclidean Distance - Measures geometric distance.
Example
Used in error detection and correction codes.
Heteroscedastic Regression
A regression model where the variance of the errors is not constant across observations.
Types of Regression Models
- Homoscedastic Regression - Assumes equal variance.
- Heteroscedastic Regression - Assumes varying variance.
Example
Used in financial market predictions.
Hierarchical Agglomerative Clustering (HAC)
A bottom-up clustering method that merges similar clusters iteratively.
Types of Hierarchical Clustering
- Agglomerative Clustering - Starts with individual points.
- Divisive Clustering - Starts with a single cluster and splits.
Example
Used in customer segmentation.
High-Dimensional Data Analysis
Techniques used to process and extract patterns from data with many features.
Types of Dimensionality Reduction
- Principal Component Analysis (PCA) - Reduces features while preserving variance.
- T-SNE - Used for visualizing high-dimensional data.
Example
Used in bioinformatics and genetics.
Hidden Markov Model (HMM)
A statistical model where the system is assumed to follow a Markov process with hidden states.
Types of Hidden Markov Models
- Discrete HMM - Uses a finite set of states.
- Continuous HMM - Uses continuous probability distributions.
Example
Used in speech recognition and natural language processing.
Hinge Loss
A loss function used in classification models, particularly in support vector machines (SVMs).
Types of Loss Functions
- Hinge Loss - Used in SVMs.
- Cross-Entropy Loss - Used in deep learning.
Example
Used in SVM-based text classification.
Histogram of Oriented Gradients (HOG)
A feature descriptor used to detect objects by analyzing gradient orientations.
Types of Feature Descriptors
- HOG - Captures edge directions.
- SIFT - Captures keypoint features.
Example
Used in pedestrian detection.
Hopfield Network
A type of recurrent neural network used for associative memory storage.
Types of Hopfield Networks
- Binary Hopfield Network - Uses discrete neurons.
- Continuous Hopfield Network - Uses continuous-valued neurons.
Example
Used in pattern recognition and error correction.
Hyperparameter Optimization
The process of selecting the best hyperparameters for a machine learning model to maximize performance.
Types of Hyperparameter Optimization
- Grid Search - Evaluates all possible hyperparameter combinations.
- Bayesian Optimization - Uses probabilistic models to find optimal parameters.
Example
Used in deep learning to tune learning rates and batch sizes.
Hyperplane
A geometric concept that separates different classes in a high-dimensional space.
Types of Hyperplanes
- Linear Hyperplane - Separates data using a straight line or plane.
- Nonlinear Hyperplane - Uses complex boundaries (e.g., kernel-based SVMs).
Example
Used in support vector machines (SVMs) for classification.
Hybrid Learning
A machine learning approach that combines multiple learning techniques to improve performance.
Types of Hybrid Learning
- Ensemble Learning - Uses multiple models for better accuracy.
- Neuro-Symbolic Learning - Combines deep learning with symbolic AI.
Example
Used in recommendation systems to combine collaborative and content-based filtering.
Hashing Trick
A technique that maps high-dimensional data into a lower-dimensional space using hash functions.
Types of Hashing Techniques
- Feature Hashing - Reduces dimensionality in text processing.
- Locality-Sensitive Hashing - Finds approximate nearest neighbors.
Example
Used in text classification for feature reduction.
Huber Loss
A loss function that combines mean squared error (MSE) and mean absolute error (MAE) to reduce sensitivity to outliers.
Types of Loss Functions
- Huber Loss - Balances MSE and MAE.
- Log-Cosh Loss - Similar to Huber but differentiable everywhere.
Example
Used in robust regression models.
Hierarchical Reinforcement Learning (HRL)
A reinforcement learning method that breaks down tasks into sub-goals for better efficiency.
Types of HRL
- Option-Based HRL - Uses predefined sub-policies.
- Goal-Based HRL - Defines specific hierarchical objectives.
Example
Used in robotics for complex task execution.
Homomorphic Encryption in ML
A cryptographic technique that allows computations on encrypted data without decryption.
Types of Homomorphic Encryption
- Partial Homomorphic Encryption - Supports limited operations.
- Fully Homomorphic Encryption - Supports arbitrary computations.
Example
Used in privacy-preserving machine learning.
Hessian Matrix
A second-order derivative matrix used to analyze curvature in optimization problems.
Types of Hessian Matrices
- Positive Definite Hessian - Indicates local minimum.
- Negative Definite Hessian - Indicates local maximum.
Example
Used in deep learning optimization algorithms like Newton’s method.
High-Order Markov Model
An extension of the Markov model where future states depend on multiple previous states.
Types of Markov Models
- First-Order Markov Model - Depends only on the previous state.
- Higher-Order Markov Model - Uses more historical context.
Example
Used in natural language processing for text prediction.
Heuristic Search
An optimization technique that uses domain knowledge to guide the search process.
Types of Heuristic Search
- Greedy Search - Expands the best candidate at each step.
- A* Search - Balances cost and heuristic estimates.
Example
Used in AI planning and game algorithms.
Histogram Equalization
A technique that improves image contrast by adjusting intensity distributions.
Types of Histogram Equalization
- Global Histogram Equalization - Adjusts the entire image.
- Adaptive Histogram Equalization - Works on small regions.
Example
Used in medical imaging for enhanced X-ray contrast.
Hybrid Neural Networks
A neural network architecture that combines different types of networks for improved performance.
Types of Hybrid Neural Networks
- CNN-RNN Hybrid - Uses CNNs for feature extraction and RNNs for sequence modeling.
- GAN-RNN Hybrid - Uses Generative Adversarial Networks with RNNs.
Example
Used in autonomous driving systems.
Hyperbolic Tangent (Tanh) Activation Function
An activation function that outputs values between -1 and 1, used in deep learning.
Types of Activation Functions
- Sigmoid - Outputs values between 0 and 1.
- ReLU - Outputs zero for negative inputs.
Example
Used in LSTMs for recurrent neural networks.
HashNet
A deep learning model that uses hashing techniques for fast similarity search.
Types of Hashing in Deep Learning
- Supervised Hashing - Learns hash codes based on labeled data.
- Unsupervised Hashing - Learns hash codes without labels.
Example
Used in image retrieval systems.
Hopcroft-Karp Algorithm
An algorithm used to find the maximum matching in bipartite graphs, useful in ML optimization.
Types of Graph Matching
- Maximum Matching - Finds the largest set of edges without common nodes.
- Minimum Vertex Cover - Covers all edges with the fewest nodes.
Example
Used in recommendation systems.
Hamiltonian Monte Carlo (HMC)
A sampling method that uses physical simulation to explore probability distributions efficiently.
Types of Monte Carlo Methods
- Metropolis-Hastings - Uses random proposals.
- Hamiltonian Monte Carlo - Uses physics-based proposals.
Example
Used in Bayesian inference for deep learning models.
Types of Hough Transform
- Standard Hough Transform - Detects straight lines.
- Generalized Hough Transform - Detects arbitrary shapes.
Example
Used in autonomous vehicle lane detection.
Hash Embeddings
A technique that reduces memory usage in NLP models by hashing word indices.
Types of Embedding Techniques
- Word2Vec - Uses neural networks for embeddings.
- FastText - Uses subword embeddings.
Example
Used in large-scale text classification.
Hierarchical Softmax
A computationally efficient alternative to standard softmax, used for large vocabulary classification.
Types of Softmax Optimization
- Hierarchical Softmax - Uses tree-based computation.
- Negative Sampling - Reduces computation by training on negative examples.
Example
Used in Word2Vec models.
Heterogeneous Computing
A computing paradigm that utilizes multiple types of processors for performance improvement.
Types of Heterogeneous Computing
- CPU-GPU Hybrid Computing - Uses both CPU and GPU.
- FPGA Acceleration - Uses Field Programmable Gate Arrays.
Example
Used in deep learning acceleration.
Hinge Loss
A loss function used for training classifiers, especially in support vector machines (SVMs).
Types of Hinge Loss
- Standard Hinge Loss - Used in linear SVMs.
- Squared Hinge Loss - A modified version with squared penalty.
Example
Used in SVM-based text classification.
Hierarchical Clustering
A clustering method that builds a hierarchy of clusters using a tree-like structure.
Types of Hierarchical Clustering
- Agglomerative Clustering - Bottom-up approach merging clusters.
- Divisive Clustering - Top-down approach splitting clusters.
Example
Used in gene sequence analysis.
Human-in-the-Loop Machine Learning
A technique where humans assist ML models in training or decision-making.
Types of Human-in-the-Loop Learning
- Active Learning - Humans label only the most useful data.
- Reinforcement Learning with Human Feedback - Humans help guide training.
Example
Used in AI-assisted medical diagnosis.
Homoscedasticity
A property in statistical models where variance remains constant across data points.
Types of Variance in ML
- Homoscedastic - Variance is constant.
- Heteroscedastic - Variance changes with input.
Example
Used in linear regression assumptions.
Heterogeneous Graph Neural Networks (HGNNs)
Graph neural networks that process multiple types of nodes and edges in graphs.
Types of Graph Neural Networks
- Homogeneous GNNs - Work with single-type nodes.
- Heterogeneous GNNs - Handle diverse node types.
Example
Used in recommendation systems.
Hopfield Networks
A type of recurrent neural network that stores patterns and retrieves them using associative memory.
Types of Hopfield Networks
- Binary Hopfield Networks - Use discrete states.
- Continuous Hopfield Networks - Use continuous states.
Example
Used in content-addressable memory systems.
Histogram of Oriented Gradients (HOG)
A feature descriptor used for object detection in computer vision.
Types of Feature Descriptors
- SIFT - Detects key points in images.
- HOG - Analyzes gradient orientation histograms.
Example
Used in pedestrian detection in self-driving cars.
Hybrid Attention Mechanisms
Attention mechanisms that combine different types of attention strategies for improved learning.
Types of Attention Mechanisms
- Self-Attention - Focuses on relevant parts within the same input.
- Cross-Attention - Attends to information from another input.
Example
Used in transformers like BERT and GPT.
Hierarchical Bayesian Models
A Bayesian approach where model parameters follow a hierarchy of distributions.
Types of Bayesian Models
- Standard Bayesian Models - Assume independent priors.
- Hierarchical Bayesian Models - Use nested priors.
Example
Used in probabilistic programming.
Hypothesis Testing in Machine Learning
A statistical method used to validate assumptions about data and model performance.
Types of Hypothesis Testing
- Parametric Tests - Assume a specific data distribution.
- Non-Parametric Tests - Do not assume a fixed distribution.
Example
Used in A/B testing for model evaluation.
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.