Naïve Bayes Classifier
A probabilistic machine learning model based on Bayes' Theorem, assuming independence among features.
Types of Naïve Bayes
- Gaussian Naïve Bayes - Assumes normal distribution of features.
- Multinomial Naïve Bayes - Used for text classification.
Example
Used in spam email classification.
Neural Networks
Computational models inspired by biological neural networks, used in deep learning.
Types of Neural Networks
- Feedforward Neural Network - Basic type with one-way connections.
- Recurrent Neural Network - Has loops allowing memory of previous inputs.
Example
Used in image recognition and speech processing.
Noise in Machine Learning
Unwanted or irrelevant variations in data that can affect model performance.
Types of Noise
- Label Noise - Incorrectly labeled data.
- Feature Noise - Errors in input features.
Example
Used in denoising autoencoders for image processing.
Normalization
A preprocessing technique used to scale data within a specific range.
Types of Normalization
- Min-Max Scaling - Scales data between 0 and 1.
- Z-Score Normalization - Centers data around mean with unit variance.
Example
Used in deep learning to improve model convergence.
Named Entity Recognition (NER)
A natural language processing (NLP) task that identifies entities such as names, dates, and locations in text.
Types of Entities
- Person - Identifies people's names.
- Organization - Recognizes companies, institutions, etc.
Example
Used in search engines and chatbots.
Nash Equilibrium
A game theory concept where no player can improve their outcome by changing their strategy unilaterally.
Types of Nash Equilibrium
- Pure Strategy Nash Equilibrium - Players choose a single strategy.
- Mixed Strategy Nash Equilibrium - Players use probabilistic strategies.
Example
Used in reinforcement learning for multi-agent environments.
Natural Gradient Descent
An optimization technique that adjusts learning based on the geometry of the parameter space.
Types of Gradient Descent
- Vanilla Gradient Descent - Uses Euclidean updates.
- Natural Gradient Descent - Considers information geometry.
Example
Used in training deep neural networks for faster convergence.
Nearest Neighbor Algorithm
A type of instance-based learning where classification is based on the closest training examples.
Types of Nearest Neighbor
- K-Nearest Neighbors (KNN) - Considers 'K' closest points.
- Weighted KNN - Weighs closer neighbors more.
Example
Used in recommendation systems and image recognition.
Negative Log-Likelihood (NLL)
A loss function used in probabilistic models to measure how well predictions match true outcomes.
Types of Likelihood Estimation
- Maximum Likelihood Estimation (MLE) - Finds parameters that maximize probability.
- Bayesian Estimation - Uses prior distributions for likelihood computation.
Example
Used in classification models such as logistic regression.
Neural Architecture Search (NAS)
An automated process to design optimal deep learning architectures.
Types of NAS
- Reinforcement Learning-Based NAS - Uses RL to explore architectures.
- Evolutionary Algorithms-Based NAS - Uses genetic algorithms.
Example
Used in AutoML frameworks to optimize deep learning models.
Noise Reduction
A technique used to remove unwanted variations in data that can negatively affect machine learning models.
Types of Noise Reduction
- Filtering - Uses statistical methods to smooth noisy data.
- Dimensionality Reduction - Removes noise by reducing feature space.
Example
Used in image and speech processing to improve data quality.
Non-Convex Optimization
A type of optimization problem where the objective function has multiple local minima.
Types of Optimization
- Convex Optimization - Single global minimum.
- Non-Convex Optimization - Multiple local minima.
Example
Used in deep learning training where loss landscapes are highly complex.
Nonlinear Activation Functions
Functions used in neural networks to introduce non-linearity, enabling complex decision boundaries.
Types of Nonlinear Activation Functions
- ReLU - Popular in deep learning.
- Tanh - Scales values between -1 and 1.
Example
Used in convolutional neural networks (CNNs) to process images.
Nonparametric Models
Machine learning models that do not assume a fixed functional form for data distribution.
Types of Nonparametric Models
- Decision Trees - Splits data into hierarchical rules.
- Kernel Density Estimation - Estimates probability densities.
Example
Used in anomaly detection and clustering.
Normal Equation
A mathematical method to compute linear regression coefficients without iterative optimization.
Types of Solutions
- Closed-Form Solution - Direct computation.
- Gradient Descent - Iterative approach.
Example
Used in simple linear regression when dataset size is small.
Normalized Discounted Cumulative Gain (NDCG)
A ranking evaluation metric that measures the quality of search engine results.
Types of Ranking Metrics
- DCG - Measures relevance while considering rank position.
- NDCG - Normalizes DCG for comparison across queries.
Example
Used in search engines like Google to evaluate ranking quality.
Novelty Detection
A technique used to identify new or unusual patterns in data that differ from known examples.
Types of Novelty Detection
- Supervised - Uses labeled training data.
- Unsupervised - Detects anomalies without prior labels.
Example
Used in fraud detection systems to spot suspicious transactions.
Null Hypothesis in ML
A statistical hypothesis stating that there is no significant difference between observed and expected data.
Types of Hypothesis Testing
- Null Hypothesis (H0) - Assumes no effect.
- Alternative Hypothesis (H1) - Assumes a significant effect.
Example
Used in A/B testing to determine if a new feature improves user engagement.
Numerical Stability
A property of algorithms ensuring that small numerical errors do not accumulate excessively.
Types of Numerical Stability
- Stable Algorithms - Errors remain controlled.
- Unstable Algorithms - Errors grow exponentially.
Example
Used in deep learning frameworks to avoid vanishing gradients.
Nystrom Method
A technique used to approximate large kernel matrices in machine learning.
Types of Kernel Approximation
- Random Fourier Features - Uses random projections.
- Nystrom Method - Uses selected data points.
Example
Used in kernel-based SVMs for large datasets.
N-gram Model
A probabilistic language model that predicts the next word in a sequence based on the previous 'N' words.
Types of N-gram Models
- Unigram - Considers only one word at a time.
- Bigram/Trigram - Considers two or three consecutive words.
Example
Used in speech recognition and predictive text input.
Neural Style Transfer (NST)
A deep learning technique that applies the artistic style of one image onto another.
Types of Style Transfer
- Single Style Transfer - Transfers one style to an image.
- Multi-Style Transfer - Blends multiple artistic styles.
Example
Used in AI-powered art applications like DeepArt.
Non-Negative Matrix Factorization (NMF)
A dimensionality reduction technique where matrix elements remain non-negative.
Types of Matrix Factorization
- Singular Value Decomposition (SVD) - Factorizes matrices with positive and negative values.
- Non-Negative Matrix Factorization (NMF) - Retains only non-negative values.
Example
Used in topic modeling and recommendation systems.
Neural Tangent Kernel (NTK)
A mathematical framework used to study the training dynamics of deep neural networks.
Types of NTK Analysis
- Finite-Width NTK - Accounts for practical network sizes.
- Infinite-Width NTK - Approximates networks as Gaussian processes.
Example
Used in theoretical deep learning research to analyze network convergence.
Node Embedding
A representation learning technique that encodes graph nodes into vector space for analysis.
Types of Node Embedding
- Random Walk-Based - Uses node co-occurrence patterns.
- Matrix Factorization-Based - Decomposes adjacency matrices.
Example
Used in social network analysis and recommendation systems.
Noisy Student Training
A semi-supervised learning approach that improves model robustness by adding noise during training.
Types of Noisy Student Training
- Standard Noisy Student - Uses labeled and unlabeled data.
- Iterative Noisy Student - Repeats the process for better generalization.
Example
Used in large-scale vision models like EfficientNet.
Neural Variational Inference
A Bayesian inference technique that approximates complex probability distributions using neural networks.
Types of Variational Inference
- Mean-Field Variational Inference - Assumes independent latent variables.
- Structured Variational Inference - Captures dependencies between variables.
Example
Used in generative models like Variational Autoencoders (VAEs).
Nesterov Accelerated Gradient (NAG)
An optimization algorithm that improves convergence by incorporating momentum with a lookahead gradient step.
Types of Momentum Optimization
- Standard Momentum - Uses past gradients for acceleration.
- Nesterov Momentum - Anticipates future gradient direction.
Example
Used in deep learning optimizers like Adam and SGD.
Non-Autoregressive Models
Machine learning models that generate outputs in parallel instead of sequentially.
Types of Sequence Generation
- Autoregressive Models - Generate one token at a time.
- Non-Autoregressive Models - Generate multiple tokens simultaneously.
Example
Used in fast text generation models like BERT.
Newton's Method in ML
An optimization technique used to find function roots and improve training efficiency.
Types of Newton's Method
- Standard Newton's Method - Uses second-order derivatives.
- Quasi-Newton Methods - Approximate second derivatives.
Example
Used in logistic regression and optimization algorithms like BFGS.
Negative Log Likelihood (NLL)
A loss function used in probabilistic models to measure the likelihood of observed data given a model.
Types of Likelihood Functions
- Maximum Likelihood Estimation - Maximizes probability of observed data.
- Bayesian Likelihood - Incorporates prior distributions.
Example
Used in classification models like softmax regression.
Nested Cross-Validation
A robust model validation technique that prevents data leakage by separating hyperparameter tuning and evaluation.
Types of Cross-Validation
- K-Fold Cross-Validation - Splits data into k subsets.
- Nested Cross-Validation - Adds an inner loop for hyperparameter tuning.
Example
Used in hyperparameter optimization to ensure unbiased performance estimation.
Neural Network Pruning
A technique for reducing the size of neural networks by removing redundant parameters.
Types of Pruning
- Weight Pruning - Removes individual weights.
- Neuron Pruning - Removes entire neurons or layers.
Example
Used to compress deep learning models for mobile applications.
No-Free-Lunch Theorem
A fundamental theorem stating that no single machine learning algorithm performs best across all tasks.
Types of No-Free-Lunch Theorems
- Optimization - No single optimizer is universally best.
- Generalization - No algorithm generalizes perfectly across datasets.
Example
Used to justify the need for algorithm selection based on specific tasks.
Neural Differential Equations
A framework that integrates deep learning with differential equations for modeling continuous-time systems.
Types of Neural Differential Equations
- ODE-based - Uses ordinary differential equations.
- PDE-based - Uses partial differential equations.
Example
Used in physics-informed neural networks.
Non-Parametric Density Estimation
A technique for estimating probability distributions without assuming a predefined functional form.
Types of Density Estimation
- Parametric - Assumes a specific distribution (e.g., Gaussian).
- Non-Parametric - Does not assume a specific distribution.
Example
Used in kernel density estimation (KDE) and histograms.
Neural Architecture Search (NAS)
An automated method for designing optimal neural network architectures.
Types of NAS
- Reinforcement Learning-Based - Uses RL to search architectures.
- Evolutionary Algorithm-Based - Uses genetic algorithms.
Example
Used in AutoML to automate deep learning model selection.
Nodal Analysis in Graph Neural Networks
A technique for analyzing and propagating information in graph-based deep learning models.
Types of Graph Analysis
- Node Classification - Assigns labels to nodes.
- Link Prediction - Predicts connections between nodes.
Example
Used in recommender systems and knowledge graphs.
Non-Stationary Time Series
A time series where statistical properties such as mean and variance change over time.
Types of Time Series
- Stationary - Mean and variance remain constant.
- Non-Stationary - Trends or seasonality affect data.
Example
Used in financial market predictions.
Nonlinear Regression
A regression technique where the relationship between variables is represented by a nonlinear function.
Types of Regression
- Linear Regression - Assumes a straight-line relationship.
- Nonlinear Regression - Uses polynomial, exponential, or logarithmic models.
Example
Used in biological and economic forecasting models.
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.