Glossary

ACID transactions are qualities that assure database transaction dependability and consistency.

Specialized team within an organization dedicated to driving the development, deployment, and strategic implementation of AI technologies and initiatives.

It is the process of recognizing and labeling data samples that are particularly crucial in supervised learning in Machine Learning.

Ensuring ethical AI use to prevent biases and promote equity across society.

Essential process to ensure AI/ML model accuracy, reliability, and security.

Critical for maintaining AI system reliability and transparency through ongoing monitoring and insights.

Mathematical activation functions are used to the outputs of artificial neurons in a neural network to make the model nonlinear.

ML technique where models query specific data for labeling to improve learning efficiency.

AdaGrad is a renowned optimization algorithm utilized extensively in ML and DL. its primary function is the alteration of the learning rate throughout the training period.

Hostile Machine Learning encompasses the methods of Machine Learning designed to generate or pinpoint adversarial instances.

An anomaly, synonymous with abnormality, deviation, or discrepancy is termed as an event or situation that deviates from ordinary happenings.

Emulates how humans learn, as suggested by the term 'neural' in their name. ANNs utilize algorithms for adaption and learning from novel data inputs.

Attributes represent the various data elements that are used. These attributes often go by other names, such as fields, features, or variables.

Neural networks that encode data to a lower dimension and decode it back, often for noise reduction or feature extraction.

A technique designed to streamline the tedious aspects of model development.

A statistical tool used for predicting future data points based on preceding ones in a particular time series.

Calculated by averaging the precision over all recall levels ranging from 0 to 1 at different IoU

Bidirectional Encoder Representations from Transformers, aids computers in comprehending ambiguous word meanings within the given text by employing nearby context.

"Bilingual Evaluation Understudy", is a measurable statistic used to evaluate the accuracy of translations produced by machines as against those made by human translators.

A computed aided mathematical method for optimizing predictions in the sphere of data mining and machine learning.

This algorithm is a common instructional process adopted by neural networks to calculate the steepest descent.

A machine learning technique aimed at increasing the stability of prediction models and reducing variance.

Baseline models serve as the initial models which assist as a foundation for the development of more intricate models.

Batch standardization or normalization is a method within the domain of deep learning that markedly enhances the performance and dependability of neural network structures.

Is a valuable instrument that allows for the computation of conditional probability.

In machine learning, balancing bias (model's prediction discrepancy) and variance (prediction variability) is vital. Too much bias can miss underlying data patterns, leading to underfitting. High variance can overfit, memorizing noise. Striking a balance helps create accurate models.

In machine learning, binary classification involves distinguishing between two classes. Examples include spam detection or disease diagnosis. It's a subset of classification tasks, which also includes multi-class and multi-label classifications. Key algorithms for binary tasks include Logistic Regression and Support Vector Machines.

Serves as a critical statistical probability distribution that details the amount of successful outcomes from a predetermined set of independent trials, each possessing identical success chances.

Refers to systems where the internal workings are unclear or hidden. While they process inputs and produce outputs, the underlying mechanics remain mysterious. Such models, although prevalent in industries like finance and healthcare, pose challenges, especially when biases creep in. Transparent counterparts, or 'white box' models, offer inspectable and understandable mechanics.

A method adapting Continuous Integration and Deployment for ML, ensuring automated, seamless model training and deployment.

Calibration in machine learning evaluates the alignment of a model's predicted probabilities with actual outcomes. Calibration curves visually depict this, aiming for predicted probabilities to reflect real-world occurrences.

Standardized data model that provides a unified representation of data across multiple systems and platforms. It serves as a "universal language" for data, ensuring consistent communication between diverse systems.

A phenomenon in Neural Networks where old knowledge is lost upon learning new information. It's akin to severe human amnesia.

A machine learning algorithm by Yandex, known for efficiently handling categorical data. It simplifies model-building without needing extensive data pre-processing.

Distinct data categories without intrinsic numeric values. Divided into nominal (unordered) and ordinal (ordered). Crucial for ML preprocessing.

An advanced NLP model optimized for dialogues, evolving from Generative Language Models (GLMs). Known for contextual comprehension, scalability, and coherent responses. Essential for AI-enhanced conversational tools.

A machine learning challenge where one class significantly outnumbers another, potentially biasing models and hindering predictive accuracy.

A decisive point in machine learning that determines the boundary between classes based on predicted probabilities or scores. It impacts model precision and recall.

Techniques in unsupervised machine learning used to group similar data points together. They identify structures or patterns in untagged datasets, forming distinct clusters.

Sophisticated method in the realm of information technology and data processing that continuously observes, identifies, and analyzes real-time data and specific events.

Specialized area of computer science that focuses on developing digital frameworks capable of processing, interpreting and understanding visual input (images or videos), in ways quite similar to human capabilities.

A tool that evaluates classifier accuracy by comparing actual vs. predicted classifications.

DevOps practice where code changes are automatically integrated, tested, and merged into the main codebase frequently, enhancing software quality and development speed.

An automated process ensuring newly developed code maintains quality and integrates seamlessly with the main repository, enhancing development speed and reducing glitches.

Software that strive to simulate human-like conversations with users through text or voice.

A special subset of optimization that emphasizes the minimization of a convex objective function while adhering to convex constraints.

A leading deep learning model predominantly employed for image processing and other computer vision tasks.

A pivotal technique in model assessment, cross-validation ensures robust performance on unseen data by testing models against multiple data subsets, enhancing prediction reliability.

Also known as data labeling, is the act of identifying and marking data samples necessary for supervised learning models in Machine Learning (ML).

A technique that enhances dataset volume and diversity by introducing minor modifications to existing data, aiding in improved model performance and addressing data scarcity in machine learning.

A process in Time Series analysis that divides data into components like level, trend, seasonality, and noise, assisting in understanding inherent patterns for better forecasting.

Granularity refers to the level of detail in data. In data science, it indicates the depth of segmentation or the precision of data categorization.

Is a specialized, streamlined portion of a data warehouse, designed to accommodate a single department or business function within an enterprise.

The process of identifying and correcting inaccurate, corrupt, improperly structured, duplicated, or missing information from a dataset.

Advanced analytics and machine learning tools that facilitate data scientists in crafting strategies and extracting meaningful insights from data.

Techniques in data science include non-linear models, support vector machines, linear regression, and pattern recognition, which analyze and transform data into actionable insights.

Tools in data science, from data mining like Weka and Pandas to data analysis tools such as KNIME and Hadoop, aid in deriving insights from raw data.

A specialized methodology for data modeling and integration, specifically designed as the foundation for creating scalable, agile, and adaptive data warehouses.

A process of tracking and managing changes to datasets, data versioning is crucial for efficient machine learning experiments. It aids in quick data product creation, error mitigation, and understanding data evolution.

A method of presenting complex data as graphical representations, aiding in recognizing trends, anomalies, and patterns.

An AI approach prioritizing quality data over models, emphasizing data labeling & management to improve machine learning outcomes.

Emphasizing the role of organized data in AI, this covers the significance of training and testing data sets in ML, and the process of transforming raw data into usable datasets for business-oriented outcomes.

A hypersurface that outlines the perimeters of distinct classifications. The decision boundary represents the area within the feature space whenever the model's prediction transitions between classes.

A sophisticated methodology that employs data analysis to facilitate swifter, more precise, and fact-informed decisions, which are more reliable than mere instincts or assumptions.

A supervised ML algorithm, Decision Trees classify or predict target variables based on decision rules from prior data. Simplifying complex decision-making processes, they branch on criteria for optimal results, providing clarity on risks and rewards.

Hierarchical structures used for classification and regression in machine learning.

A kind of deep learning technology that is essentially self-taught, operating without any externally-sourced data to make its calculations.

A subset of AI that mimics the brain with neural networks to process complex data.

An advanced AI technique combining deep learning and reinforcement learning. It dynamically adjusts behaviors to maximize rewards through continuous feedback.

Advanced techniques in AI, built on neural networks. Includes CNNs for image recognition, GANs for data generation, and LSTMs for sequence data.

An ML model's performance can deteriorate over time due to concept drift, where statistical characteristics of the target variable change. Regular updates and monitoring are crucial to maintain accuracy.

A CNN architecture that distinctively links each layer to every other, enhancing learning by reusing features.

Density-based clustering utilizes machine learning methods without supervision to track down distinct clusters within a dataset.

A method in ML to simplify data by reducing redundant features, improving visualization, and enhancing model efficiency.

A premier tool in R for data manipulation. Mastery of dplyr can significantly reduce the time data scientists spend on data handling and preparation, making their tasks more understandable.

A crucial aspect of MLOps, it involves tracking deviations in ML model's predictions due to changes in real-world conditions or data integrity issues.

A systemized process to extract, transform, and load data, enabling efficient data analytics and decision-making.

a technique to curb the "overfitting" of a machine learning model. In this approach, data is split into a training set and a validation set. While the training set educates the model, the validation set assesses its progress.

A comprehensive approach to machine learning, aiming at enhancing the predictive performance by amalgamating decisions from multiple models.

An innovative branch of AI focused on creating content and designs. It aids in marketing, product design, data augmentation, and personalizing user experiences.

A vital hyperparameter that determines the number of times the complete training dataset goes through the machine learning algorithm.

AI practices aligning with ethical standards, ensuring fairness, transparency, and respecting human rights.

an advanced optimization technique that harnesses concepts of natural evolution to tailoring solutions to problems concerning function optimization.

AI designed to be transparent and provide insights into its decision-making, fostering trust, understanding, and compliance.

A foundational step in the data analysis pipeline. Its primary aim is to unveil the underlying structure of a dataset, often using visual tools.

a powerful tool used to gauge the performance of a Machine Learning model. It amalgamates precision and recall into a solitary score.

Through advanced biometric authentication methods, this genre of AI can determine and authenticate a person's identity by analyzing their facial features from videos or images.

Examining potential pitfalls in deploying ML models, including overlooked performance bias, data pipeline-induced model malfunctions, and robustness failures.

A metric in ML indicating the proportion of negative cases wrongly classified as positive.

The process of creating, transforming, extracting, and selecting optimal variables to enhance machine learning models' performance.

The process of identifying and including essential variables in ML algorithms to improve model performance and reduce complexity.

A n-dimensional array of numerical values representing an object's measurable attributes, pivotal in pattern recognition and machine learning for object representation.

A decentralized machine learning approach, allowing devices to learn from local data, enhancing privacy and real-time predictions.

A mechanism in ML where systems use results to refine future outputs, enhancing accuracy but potentially causing ethical concerns.

A bell-curved statistical model, pivotal in representing natural occurrences and data analysis, defined by mean and standard deviation.

A statistical probability model that posits that data points can be derived from different Gaussian distributions, each identified by its mean and covariance matrix.

A statistical method modeling the relationship between variables, allowing non-normal distributions. Adapts to various response variable distributions.

Machine learning models where two neural networks (generator & discriminator) compete to improve results. Widely used in image generation and modification.

A machine learning technique using decision trees in sequence, correcting prior errors. Boosts weak models to strong ones.

An optimization algorithm used in machine learning to minimize cost by adjusting model parameters via the gradient's steepest decline.

A method for hyperparameter tuning where all possible combinations of predefined hyperparameter values are evaluated to find the best model.

The definitive and verified data used as a benchmark in supervised learning to train and evaluate models. Essential for accurate algorithm training and performance validation.

Unusual data points that deviate significantly from others, influencing statistical results.

Data structures that map keys to values, optimizing lookups and insertions through hashing techniques.

A specific subset of data deliberately withheld during a machine learning model's training process.

An approach that combines human expertise with ML algorithms to improve accuracy, especially in complex or sensitive tasks.

A process in machine learning to find the best hyperparameter settings that improve model accuracy. This involves techniques like grid, random, and Bayesian search.

In machine learning, a decision boundary that separates data into classes. Commonly used in classification algorithms like SVMs, it divides the input space optimally.

A method to artificially increase dataset size by altering original images, aiding neural networks in enhancing accuracy and generalization.

Gathering and organizing data, primarily images, for training machine learning models in computer vision tasks.

Refers to uneven representation of classes in datasets. Addressing it ensures accurate model evaluation and better prediction outcomes in classification problems.

Refers to data where each instance is random and maintains the same statistical properties. It's a foundational concept in data science ensuring reliable statistical inference.

A technological solution aimed at sorting structured data from various documents like contracts, invoices, and purchase orders.

A supervised ML algorithm that classifies data based on the majority vote of its 'K' closest labeled data points.

A non-parametric, lazy learning algorithm used for classification and regression. It classifies based on the 'K' closest labeled data points.

A verification method used by companies, especially in the financial sector, to identify their clients and deter illegal activities.

Advanced AI models skilled in understanding and generating language, serving as dynamic interfaces for user interactions.

A tool designed for diagnosing and refining Large Language Models, offering insights and optimization solutions for AI practitioners.

A multifaceted approach to assess Large Language Models, ensuring their accuracy, fairness, and ethical compliance in AI applications.

Instances where AI, specifically Large Language Models, produce outputs deviating from input's reality, highlighting complexities in AI development and behavior.

The foundational elements that shape AI behavior, LLM parameters determine how these models interpret language and craft outputs, crucial for optimal AI performance.

An exploration of how Large Language Models (LLMs) are revolutionizing the domain of data summarization, offering swift, unbiased solutions.

A specialized approach focusing on deploying and managing low latency or real-time machine learning models, enhancing prediction speeds and inference performance.

A pioneering AI language model known for its superior text comprehension and generation, offering advanced features for diverse applications, from chatbots to sentiment analysis.

A hyper-parameter that controls how much a model adjusts in response to the estimated error each time the model weights are updated.

A gradient boosting framework that uses tree-based algorithms, prioritizing speed and efficiency. Known for leaf-wise tree growth to optimize loss.

A statistical method used to model and analyze the relationships between a dependent and one or more independent variables. Central to predictive analytics.

A groundbreaking Large Language Model Architecture excelling in NLP tasks. Renowned for its intricate design and vast parameter span, LLama interprets and produces human-like text.

It provides insights into the decision-making process of any machine learning model, whether it be neural networks, decision trees, or support vector machines, hence the term "model-agnostic"

A statistical method used in predictive modeling, where the outcome is categorical. It helps understand relationships between variables and predict binary or multinomial outcomes, aiding in data-driven decision-making processes.

A type of Recurrent Neural Network (RNN) that retains sequence data and long-term patterns effectively. Common in deep learning tasks.

An AI methodology enhancing AI model performance and flexibility by simplifying customization using low-rank techniques.

The blueprint for developing ML systems, focusing on data management, model building, and predictions. Essential for scalable AI solutions.

A systematic approach to identify and rectify issues in ML models, ensuring optimized performance. Vital for model reliability and accuracy.

The foundational setup and tools used in building, training, and deploying ML models. Essential for scalable and efficient machine learning operations.

The ability to understand and describe the decision-making process of ML models, focusing on how the models make predictions.

A standardized documentation tool for ML models, highlighting training, performance, biases, and intended applications.

A system that oversees the ML model lifecycle, ensuring efficient tracking, versioning, deployment, and monitoring of machine learning models.

A crucial process in ML that ensures models perform accurately and efficiently on their intended data, reducing risks like overfitting before deployment.

Refers to the capability of a machine learning system or model to process enormous amounts of data or manage substantial traffic without compromising on performance or precision.

A collection of software tools and frameworks designed to facilitate the development and deployment of machine learning solutions, covering data preparation, modeling, deployment, and analysis.

Combining ML development and operations to ensure efficient deployment, management, and scaling of ML systems.

A methodology combining ML development and operations, streamlining ML lifecycles, enhancing collaboration, and addressing challenges from data unpredictability in deployment.

A critical process in the ML lifecycle to maintain model relevance, track performance metrics, and establish feedback loops, ensuring consistent and robust ML system functionality.

Harnessing MLOps in the domain of Generative AI to streamline and optimize the development, deployment, and maintenance of creative AI applications.

A subset of AI that allows systems to learn and improve from data, enabling tasks such as prediction, classification, and recommendation.

An AI technique enabling computers to learn from data and make decisions without explicit coding.

An error in training data causing ML models to make skewed predictions. Reflects societal or human biases in algorithms, requiring careful mitigation.

A technique to periodically save interim model states during training. Aids in recovery from disruptions, conserves resources, and facilitates model performance tracking.

Integrating ML models into real-world applications, often interfacing with APIs. Critical for operational advantages.

The phase of executing an ML model on real-time data to generate predictions, distinguishing between training and practical application.

A structured approach to ML projects, encompassing stages from data collection to deployment, ensuring models address the identified problems efficiently.

A measure of an ML model's prediction correctness. It's vital for decision-making but isn't the sole metric; precision & recall also matter.

A process assessing the effectiveness of ML models. Involves accuracy metrics and performance metrics like AUC, F1-score.

Refers to the process of automating the deployment, management, and monitoring of machine learning models at scale.

A structured framework facilitating ML model creation. Encompasses data preparation, model training, deployment, and monitoring to optimize ML operations.

A systematic approach guiding ML model development, encompassing data collection, preprocessing, dataset creation, refinement, training, evaluation, and potential challenges like data cleanliness and concept drift.

A cloud service offering machine learning tools without the need for in-house development. Ideal for enhancing business processes and strategies.

Leveraging ML to enhance software testing processes, boosting precision, efficiency, and reliability.

A metric for regression model accuracy. It measures the average size of prediction errors, disregarding their direction.

A risk function quantifying the average of squared differences between actual and predicted values. Lower MSE indicates better accuracy.

A machine learning approach where algorithms refine and train other ML models, enabling AI systems to generalize knowledge across tasks and acquire new capabilities swiftly.

Absent data points in sequential datasets. Vital to manage, either by removal or suitable substitution, to ensure accurate modeling and analysis in time series data.

The mechanism through which an ML model operates and predicts. Influenced by training data, structure, and procedures, it requires consistent monitoring to ensure accuracy, fairness, and efficiency in predictions.

Fine-tuning predictions to align expected probabilities of a model with real-world outcomes, enhancing accuracy and trust.

Shift in data patterns over time, affecting model accuracy. Continuous monitoring and refinement are essential for maintaining performance.

Ensuring AI & ML models predict without discrimination, considering factors like age, race, or gender. Vital for equitable and just outcomes.

Observing and measuring ML model performance in real-time to detect issues and ensure it meets business requirements consistently. Vital for MLOps.

The practice of examining and understanding the inner workings and performance of ML models in operational settings. Essential for ensuring robustness, reliability, and optimizing performance over time. Utilizes tools for monitoring, logging, visualization, and analysis.

Internal coefficients of ML models adjusted during training to improve accuracy.

A centralized storage system for ML models, streamlining lifecycle management, reducing data loss, and ensuring smooth collaboration between teams.

The process of updating an existing ML model using a new dataset to improve its performance or adapt to new tasks, ensuring accuracy and efficiency over time.

The procedure of choosing the most fitting machine learning model after evaluating multiple models. Essential for balancing bias and variance, and employing resampling techniques like K-Fold Cross-Validation ensures optimal generalization to unseen data.

Adjusting hyperparameters to optimize model performance. Helps in refining the learning process and improving predictive accuracy.

A method tailoring solutions for specific problems using Factor graphs, Bayesian inference, and Probabilistic Programming.

A software development approach focusing on platform-independent models, offering abstraction and flexibility in crafting software solutions.

A machine learning task where instances are categorized into one of several predefined classes. Unlike binary classification, which has two outcomes, multi-class can have multiple outcomes, but each instance belongs to just one class.

An artificial neural network with multiple layers. Processes complex data through input, hidden, and output layers. Effective for classification and regression tasks.

A probabilistic classifier based on Bayes' Theorem with an assumption of independence among predictors. Widely used in text categorization and spam detection.

A subfield of AI focused on interpreting and comprehending human language. Enables machines to interact, analyze, and respond to human language inputs.

The optimization process of hyperparameters in a neural network to enhance performance. This involves adjusting layers, neuron counts, learning rate, batch size, and loss functions.

Software constructs that emulate human neural activity, crucial for tasks like machine learning and natural language processing.

Platforms that simplify machine learning processes, allowing users to build apps and integrations without manual coding. Enhances AI capabilities and automations.

Unwanted fluctuations in data, which can distort analysis and lead to erroneous model generalizations. Requires techniques like PCA for mitigation.

Refers to unwanted variations in color or brightness in images, often due to sensor limitations or conditions during capture. Filters are used for denoising.

A process of rescaling data to a standard range, often used when feature ranges vary. Two main types are Min-Max and Standardization Scaling. It helps in faster convergence and accurate predictions in certain algorithms.

Conversion of categorical data into binary columns for easier processing by machine learning models.

Tools and systems to oversee and manage ML models in operation, ensuring optimal performance, accuracy, and compliance.

In AI, out-of-distribution (OOD) refers to data that differs substantially from training data.

When a model learns training data too well, including its noise and outliers, reducing its ability to generalize to new data.

Essential Python libraries for data manipulation and scientific computation. While Pandas focuses on data analysis using tables, NumPy emphasizes numerical operations with arrays.

A benchmark dataset for object detection and segmentation. It offers a standardized evaluation for object detection methods with detailed annotations using bounding boxes.

A technique used to identify and manage specific patterns in data. Essential for code validation and data analysis.

In CNNs, pooling layers compress output data from convolutional layers, preventing overfitting and reducing computational needs.

Precision quantifies how correctly positive outcomes are predicted.

A process to verify a predictive model's accuracy using separate training and testing datasets.

The act of transforming raw data into a usable format, addressing inconsistencies, missing values, and noise. Essential for Data Mining.

PCA is a method that reduces data dimensions, preserving core information, crucial for machine learning.

A machine learning method that estimates the likelihood of data belonging to various classes, rather than a definitive class prediction.

The craft of refining prompts to guide AI language models, like GPT-3, towards desired responses, enhancing their accuracy and applicability.

An initial, simplified version of a machine learning model built to test its functionality, efficiency, and get early feedback. It facilitates swift iterations, saving time and resources.

A deep learning platform born from Torch, known for its tensor library, GPU support, and Pythonic design. Ideal for beginners in neural networks.

An optimization algorithm used in deep learning to speed up and stabilize model training by adjusting learning rates based on the moving average of squared gradients.

An ensemble machine learning technique using multiple decision trees for improved accuracy and reduced overfitting in classification and regression tasks.

A technique to initialize neural network weights with random values close to zero, ensuring diverse neuron outputs and aiding efficient gradient descent.

A metric that measures the ratio of correctly predicted positive observations to the actual positives, highlighting a model's capability to find all relevant cases in a dataset.

Visually demonstrates the predictive ability of a binary classifier system

A popular activation function in deep learning, ReLU outputs the input if positive, otherwise zero.

Neural networks with memory, RNNs process sequences by iterating through elements and maintaining state. Used in tasks like language modeling.

A machine learning technique for predicting a continuous value. Linear Regression, a common type, estimates relationships between variables to forecast outcomes.

Tools in machine learning and statistics used to predict continuous values. They range from simple linear regression for two variables to complex methods like SVM and LASSO.

Techniques in machine learning that prevent overfitting by adjusting model parameters, promoting genuine understanding.

A technique that reduces overfitting by constraining model parameters, ensuring models generalize well to new data.

A type of machine learning where an agent learns by interacting with an environment, receiving rewards or penalties for actions, aiming to maximize long-term rewards.

Ensuring that AI research results can be reliably duplicated by others, emphasizing consistency, proper documentation, and transparency for validation and further innovation in the field.

A deep learning architecture with "skip connections" to combat vanishing gradient and train deeper neural networks for improved computer vision tasks.

An approach ensuring AI's ethical, transparent, and safe deployment, emphasizing fairness, accountability, and societal impact for beneficial tech outcomes.

A technique in linear regression that addresses multicollinearity by adding a penalty to the coefficients, improving model reliability and accuracy.

Automates repetitive tasks by mimicking human-computer interactions, enhancing business efficiency and accuracy.

RMSE is a metric that measures the average magnitude of prediction errors, with lower values indicating better predictive accuracy. Useful in regression analysis.

A system where proxy IP addresses are automatically rotated, ensuring web scraping or browsing remains undetected by restricting IP bans. Ideal for anonymous web activity.

A Python-based library offering simple and efficient tools for predictive data analysis. Renowned for machine learning, it provides algorithms for classification, regression, clustering, and more.

Machine learning revolutionizes customer segmentation, automating the grouping based on behavior & characteristics for optimized marketing.

A middle ground between supervised & unsupervised learning. It leverages both labeled & unlabeled data for training, optimizing labeling efforts.

Metrics to assess an ML model's accuracy. Sensitivity gauges true positive rate, while specificity measures true negative rate. They play pivotal roles in evaluating model performance.

A technique to clarify model predictions by determining each feature's significance. Shapley values measure the average difference in predictions with and without the feature.

A method in ML to determine the significance of each feature in a model's prediction. Helps in interpreting complex models.

A ML approach that uses labeled data to predict outcomes. Relies on historical data for insights and prediction refinement.

A supervised learning method adept at classification and regression. SVMs find optimal hyperplanes to separate classes and can handle diverse datasets.

A computational method used to approximate complex simulations. It captures the relationship between inputs and outputs without directly executing the actual simulation, saving time and computational resources.

Artificially generated data, often used in machine learning when real-world data is limited or restricted. It aids in model training, testing, and validation.

A process using algorithms to fabricate data, often for machine learning, testing, or privacy protection. It mimics real data without revealing individual specifics.

Organized data in rows & columns, resembling spreadsheets. Used widely in businesses, it traditionally employed methods like random forests, but deep learning is emerging.

A prominent open-source library for deep learning, TensorFlow is designed by Google for complex numerical computations and supports both CPUs and GPUs.

Data subset used to evaluate a final model's performance on new, unseen examples.

Measures how often the top prediction by a classifier is incorrect.

Difference between training data & real-world serving data, affecting model accuracy.

Neural networks using self-attention to process sequential data like text.

A model excelling at processing sequences like text via attention mechanisms, enabling contextual understanding.

Algorithms that use decision trees for classification or regression tasks.

A comparison tool used in machine learning algorithms, which operates by contrasting a base input, a positive input, and a negative input.

Measures the proportion of actual positives correctly identified, vital for model accuracy in classification tasks.

Incorrectly rejecting a true null hypothesis; also known as a false positive.

Failing to reject a false null hypothesis; also known as a false negative.

A model too simple to capture patterns in data, poor on both training and new data.

Models that infer patterns from untagged data without predefined answers.

Deep CNN known for its simplicity and depth, excelling in image recognition tasks.

A subset of data used to tune a model's hyperparameters and prevent overfitting.

Optimized gradient boosting library for speed and performance in ML tasks.

YOLO is a fast, accurate algorithm that detects objects in real-time by looking at images only once.

Machine learning models recognize unseen objects using related knowledge.