T-distributed Stochastic Neighbor Embedding (t-SNE)
A nonlinear dimensionality reduction technique for visualizing high-dimensional data.
Types of t-SNE
- Standard t-SNE - Optimized for visualization.
- Barnes-Hut t-SNE - Faster approximation for large datasets.
Example
Used to visualize clusters in high-dimensional data.
Tabular Data
Data structured in rows and columns, typically used in databases and spreadsheets.
Types of Tabular Data
- Structured Data - Well-defined schema (e.g., SQL databases).
- Unstructured Data - Mixed formats requiring preprocessing.
Example
Used in business analytics and customer databases.
Teacher-Student Learning
A knowledge distillation approach where a large model (teacher) trains a smaller model (student).
Types of Teacher-Student Learning
- Soft Targets - Uses probabilistic outputs for training.
- Feature Distillation - Transfers intermediate representations.
Example
Used in compressing deep learning models for mobile applications.
Temporal Difference Learning
A reinforcement learning method that updates estimates based on differences between consecutive predictions.
Types of Temporal Difference Learning
- TD(0) - Updates immediately after each step.
- TD(λ) - Uses a weighted average of past updates.
Example
Used in game-playing AI like AlphaGo.
Tensor Decomposition
A technique for breaking down multi-dimensional arrays (tensors) into simpler components.
Types of Tensor Decomposition
- CP Decomposition - Factorizes a tensor into rank-one components.
- Tucker Decomposition - Generalizes PCA to tensors.
Example
Used in recommendation systems and signal processing.
TensorFlow
An open-source machine learning framework developed by Google.
Types of TensorFlow Models
- Sequential Models - Layer-by-layer model structure.
- Functional API - Flexible model architectures.
Example
Used for deep learning, computer vision, and NLP.
Term Frequency-Inverse Document Frequency (TF-IDF)
A statistical measure used to evaluate word importance in a document relative to a collection.
Types of TF-IDF
- Unigram TF-IDF - Uses single-word frequency.
- N-gram TF-IDF - Uses sequences of words.
Example
Used in search engines and text mining.
Thompson Sampling
A reinforcement learning technique used in multi-armed bandit problems.
Types of Thompson Sampling
- Bernoulli Thompson Sampling - Used for binary rewards.
- Gaussian Thompson Sampling - Used for continuous rewards.
Example
Used in online advertising and recommendation systems.
Time Series Analysis
A technique for analyzing data points collected over time to detect trends and patterns.
Types of Time Series Analysis
- ARIMA (AutoRegressive Integrated Moving Average) - Uses past data for forecasting.
- Exponential Smoothing - Averages past observations with decreasing weights.
Example
Used in stock market prediction and weather forecasting.
Transfer Learning
A technique where a model trained on one task is adapted for a different but related task.
Types of Transfer Learning
- Inductive Transfer - Adapts knowledge to new domains.
- Transductive Transfer - Uses unlabeled data from the target domain.
Example
Used in adapting pre-trained deep learning models like BERT and ResNet.
Training Data
The dataset used to train machine learning models by adjusting weights and parameters.
Types of Training Data
- Supervised Training Data - Labeled data with input-output pairs.
- Unsupervised Training Data - Unlabeled data used for pattern recognition.
Example
Image datasets like MNIST for training digit recognition models.
Types of Transformer Models
- BERT (Bidirectional Encoder Representations from Transformers) - Used for NLP tasks.
- GPT (Generative Pre-trained Transformer) - Used for text generation.
Example
Used in machine translation and text summarization.
Tree-based Models
A class of algorithms that use decision trees to make predictions.
Types of Tree-based Models
- Decision Trees - Simple tree structures for classification.
- Random Forests - Ensembles of decision trees for better accuracy.
Example
Used in credit scoring and fraud detection.
Triangular Kernel
A kernel function used in non-parametric density estimation.
Types of Kernel Functions
- Gaussian Kernel - Uses a normal distribution.
- Epanechnikov Kernel - Uses a parabolic shape.
Example
Used in kernel density estimation for probability modeling.
Truncated SVD (Singular Value Decomposition)
A dimensionality reduction technique for decomposing matrices.
Types of SVD
- Full SVD - Decomposes a matrix into full components.
- Truncated SVD - Keeps only the most important components.
Example
Used in Latent Semantic Analysis (LSA) for text processing.
Trust Region Policy Optimization (TRPO)
A reinforcement learning algorithm for optimizing policies while maintaining stability.
Types of Policy Optimization
- TRPO - Uses a constraint-based optimization approach.
- PPO (Proximal Policy Optimization) - A simplified alternative to TRPO.
Example
Used in robotics and autonomous vehicle navigation.
Turing Test
A test to determine whether a machine exhibits human-like intelligence.
Types of AI in the Turing Test
- Weak AI - Passes the test but lacks true understanding.
- Strong AI - Hypothetically possesses true intelligence.
Example
Used to evaluate chatbots like ELIZA and GPT models.
Twin Delayed Deep Deterministic Policy Gradient (TD3)
An improved reinforcement learning algorithm that reduces overestimation bias.
Types of Policy Learning
- DDPG (Deep Deterministic Policy Gradient) - A baseline actor-critic method.
- TD3 - Improves stability with delayed updates.
Example
Used in robotic control and game-playing AI.
Two-Stage Clustering
A clustering technique that applies two sequential methods for better results.
Types of Two-Stage Clustering
- Hierarchical + K-Means - First hierarchical clustering, then K-means.
- Density-Based + Centroid-Based - DBSCAN followed by centroid-based clustering.
Example
Used in customer segmentation and image analysis.
Type I and Type II Errors
Errors that occur in hypothesis testing in statistical learning.
Types of Errors
- Type I Error - False positive (incorrectly rejecting the null hypothesis).
- Type II Error - False negative (failing to reject a false null hypothesis).
Example
Used in medical diagnostics and fraud detection.
Temporal Difference Learning
A reinforcement learning technique that updates value estimates based on observed rewards and future predictions.
Types of Temporal Difference Learning
- TD(0) - Updates based only on the immediate next state.
- TD(λ) - Uses a weighted combination of multiple future states.
Example
Used in game-playing AI like AlphaGo.
Tensor Decomposition
A mathematical technique for breaking down high-dimensional data tensors.
Types of Tensor Decomposition
- CP Decomposition - Factorizes a tensor into rank-one components.
- Tucker Decomposition - Produces a core tensor with factor matrices.
Example
Used in recommendation systems and neuroscience data analysis.
TensorFlow
An open-source deep learning framework developed by Google for building ML models.
Types of TensorFlow APIs
- TensorFlow Core - Provides low-level control.
- TensorFlow Keras - High-level API for rapid prototyping.
Example
Used in image classification and natural language processing (NLP).
Term Frequency-Inverse Document Frequency (TF-IDF)
A statistical measure used in NLP to determine the importance of words in a document.
Types of TF-IDF Variants
- Raw Frequency TF - Uses basic term counts.
- Log Normalization TF - Adjusts frequencies with logarithmic scaling.
Example
Used in search engines and spam filtering.
Testing Data
The dataset used to evaluate the performance of a trained machine learning model.
Types of Testing Data
- Hold-out Testing Data - A separate set used after training.
- Cross-validation Testing Data - Used in k-fold validation for better evaluation.
Example
Used in model evaluation for image recognition systems.
Theano
An open-source numerical computation library optimized for deep learning.
Types of Theano Features
- Symbolic Differentiation - Automates gradient calculations.
- GPU Acceleration - Enables faster computations.
Example
Used in deep learning frameworks like Keras.
Thompson Sampling
A Bayesian reinforcement learning technique for solving multi-armed bandit problems.
Types of Bandit Algorithms
- ε-Greedy - Selects random actions with probability ε.
- Upper Confidence Bound (UCB) - Prioritizes actions with high uncertainty.
Example
Used in online advertising and A/B testing.
Thresholding
A technique used in image processing and classification to separate values into different categories.
Types of Thresholding
- Global Thresholding - Uses a fixed threshold for the entire dataset.
- Adaptive Thresholding - Adjusts thresholds dynamically based on local statistics.
Example
Used in OCR and edge detection algorithms.
Tokenization
A preprocessing step in NLP that splits text into smaller units like words or subwords.
Types of Tokenization
- Word Tokenization - Splits text into words.
- Subword Tokenization - Breaks words into smaller subunits.
Example
Used in NLP models like BERT and GPT.
Topological Data Analysis (TDA)
A method of analyzing high-dimensional data using topology.
Types of TDA Techniques
- Persistent Homology - Identifies topological features across different scales.
- MAPPER Algorithm - Constructs graphs for data visualization.
Example
Used in genomics and network analysis.
Transfer Learning
A machine learning technique where a model trained on one task is adapted for another related task.
Types of Transfer Learning
- Inductive Transfer - Transfers knowledge to a different but related task.
- Transductive Transfer - Adapts a model trained in one domain to another domain.
Example
Using pre-trained ResNet for medical image classification.
Types of Transformer Models
- Encoder-Decoder Transformer - Used in translation tasks.
- Decoder-Only Transformer - Used in text generation (e.g., GPT).
Example
Used in models like BERT, GPT, and T5.
T-SNE (t-Distributed Stochastic Neighbor Embedding)
A dimensionality reduction technique for visualizing high-dimensional data.
Types of Dimensionality Reduction
- PCA - Uses linear transformations.
- T-SNE - Uses probabilistic similarity measures.
Example
Used to visualize word embeddings and image feature representations.
Twin Neural Networks
A type of neural network architecture where two networks share weights and process two inputs simultaneously.
Types of Twin Neural Networks
- Siamese Network - Used for similarity learning.
- Triplet Network - Used for ranking-based learning.
Example
Used in facial recognition and signature verification.
Time Series Forecasting
Predicting future values based on past temporal data.
Types of Time Series Models
- ARIMA - Uses statistical methods.
- LSTMs - Uses deep learning techniques.
Example
Used for stock price prediction and weather forecasting.
Top-Down and Bottom-Up Learning
Two different approaches in machine learning model training.
Types of Learning Approaches
- Top-Down Learning - Starts with a broad understanding and refines details.
- Bottom-Up Learning - Builds knowledge from individual observations.
Example
Top-down is used in supervised learning, while bottom-up is used in reinforcement learning.
Trapezoidal Rule in ML
A numerical integration method used for approximating area under a curve.
Types of Integration in ML
- Trapezoidal Rule - Approximates area using trapezoids.
- Simpson’s Rule - Uses parabolic approximations.
Example
Used in computing AUC-ROC curves for model evaluation.
Tree-Based Models
A category of machine learning models that use decision trees for predictions.
Types of Tree-Based Models
- Decision Trees - Single tree-based classification/regression.
- Random Forests - Uses multiple trees to improve predictions.
Example
Used in fraud detection and medical diagnoses.
Turing Test
A test proposed by Alan Turing to evaluate a machine’s ability to exhibit human-like intelligence.
Types of AI Evaluation Tests
- Turing Test - Measures human-like responses.
- Winograd Schema Challenge - Tests commonsense reasoning.
Example
Used as a benchmark for AI conversational agents.
Tweedie Distribution in ML
A type of probability distribution used in machine learning for modeling non-negative responses.
Types of Tweedie Models
- Tweedie Regression - Used for insurance risk modeling.
- Compound Poisson-Gamma - Used in actuarial sciences.
Example
Used in modeling claim sizes in insurance predictions.
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.