Glossary

Activation Function: A mathematical function used in neural networks to introduce non-linearity into the model, allowing it to learn more complex patterns in the data.
Actor-Critic: A type of reinforcement learning algorithm that combines an actor network that learns to select actions with a critic network that learns to estimate the expected future reward.
Adversarial Attacks: A type of security threat to machine learning models, where an attacker intentionally crafts inputs that can fool the model into making incorrect predictions or decisions
Adversarial Examples: Inputs to a machine learning model that have been intentionally designed to cause the model to make a mistake, often by adding small perturbations to the input data.
Adversarial Training: A defense mechanism against adversarial attacks, where the machine learning model is trained on adversarially generated examples to improve its robustness and generalization.
Artificial General Intelligence: A theoretical form of AI that would have the same intelligence and problem-solving ability as a human being.
Artificial General Intelligence (AGI): A hypothetical level of artificial intelligence that can perform any intellectual task that a human can do, with the same level of flexibility, creativity, and adaptability.
Artificial Intelligence: The development of computer systems to perform tasks that typically require human intelligence, such as visual perception, speech recognition, decision-making, and language translation.
Artificial Intelligence as a Service: A cloud-based service that allows developers to integrate AI capabilities into their applications without having to build the infrastructure,themselves.
Artificial Intelligence Platform: A software platform that provides tools and resources for developers to build, test, and deploy AI applications.
Artificial Life (ALife): A field of study that aims to create artificial systems that exhibit properties of living organisms, such as self-replication, adaptation, and evolution, by using techniques from artificial intelligence, robotics, and biology.
Artificial Neural Network: A type of neural network that is designed to recognize patterns and relationships in data.
Artificial Neural Network (ANN): A type of neural network that is used in machine learning and artificial intelligence, consisting of multiple interconnected nodes or neurons.
Autoencoder: A type of neural network used in unsupervised learning to learn a compressed representation of the input data.
Autoencoder: A type of neural network architecture that learns to reconstruct the input data from a compressed representation, by training an encoder network to encode the input data and a decoder network to decode the representation.
Backpropagation: A technique used in training neural networks to adjust the weights of the connections between neurons, based on the error between the predicted and actual outputs.
Bias: A systematic error in a machine learning model that results in inaccurate predictions due to an incorrect assumption or limitation in the data.
Bias: A systematic error in a machine learning model that results from training data that is not representative of the population or that contains biases or stereotypes.
Capsule Networks: A type of neural network architecture that uses groups of neurons, called capsules, to represent objects and their pose in a hierarchical manner, allowing it to handle deformation, occlusion, and other complex transformations.
Chatbot: A computer program that simulates conversation with human users, often used for customer service and support.
Classification: A technique used in supervised learning to predict a categorical output variable based on input variables.
Clustering: A technique used in unsupervised learning to group similar data points together based on their characteristics.
Clustering: A type of unsupervised learning where the algorithm groups similar data points together based on their characteristics.
Computer Vision: The ability of computers to interpret and understand visual information from the world around them.
Convolutional Generative Adversarial Networks (CGANs): A type of generative model that uses convolutional neural networks as the generator and discriminator networks in a GAN architecture, allowing it to generate realistic images and other structured data.
Convolutional Neural Network: A type of neural network that is designed to process and analyze visual data such as images and videos.
Convolutional Neural Network (CNN): A type of neural network that is commonly used for image and video recognition tasks, due to its ability to extract features from local neighborhoods of the input data.
Convolutional Neural Network (CNN): A type of neural network architecture that is specialized for processing images and other structured data, by using convolutional layers to extract local features and pooling layers to reduce dimensionality.
Curriculum Learning: A type of machine learning where the training data is presented in a sequence of increasing difficulty, to help the model learn more effectively.
Data Augmentation: The process of generating new training data from existing data by applying transformations such as rotation, scaling, and cropping.
Data Mining: The process of discovering patterns and relationships in large datasets using machine learning techniques.
Data Normalization: The process of transforming data to have a standard scale and range, often used to improve the accuracy of a machine learning model.
Data Preprocessing: The process of cleaning and transforming raw data to make it suitable for machine learning algorithms.
Decision Tree: A type of supervised learning algorithm that is used for classification and regression tasks.
Deep Belief Network (DBN): A type of neural network architecture that consists of multiple layers of RBMs, with the first layer receiving the input data and the subsequent layers forming a hierarchical representation of the input.
Deep Dream: A visualization technique that uses a neural network to enhance and exaggerate the patterns in an image, creating surreal and psychedelic effects.
Deep Learning: A subset of machine learning that uses neural networks with many layers to analyze data and improve accuracy over time.
Deep Learning: A type of machine learning that involves neural networks with many layers, capable of learning complex patterns in data.
Deep Q-Network (DQN): A type of reinforcement learning that uses a deep neural network to represent the Q-function, and uses experience replay, target networks, and other techniques to stabilize and improve its performance, by reducing the correlation between samples and the target and by preventing overfitting and instability.
Deep Reinforcement Learning: A type of machine learning that combines deep neural networks with reinforcement, learning, enabling agents to learn complex behaviors from high-dimensional input spaces.
Deep Reinforcement Learning: A type of reinforcement learning that uses deep neural networks to represent the agent's policy or value function, enabling it to learn more complex and high-dimensional tasks.
Domain Adaptation: A type of transfer learning where a model is trained on a source domain, and then adapted to a target domain with different characteristics.
Ensemble Learning: A technique in machine learning that combines multiple models to improve accuracy and reduce overfitting.
Ethical AI: The practice of developing artificial intelligence that is aligned with ethical and moral principles, such as transparency, fairness, and accountability.
Expert Systems: An AI system that is designed to replicate the decision-making abilities of a human expert in a particular domain.
Explainable AI (XAI): A field of study focused on developing machine learning models that can provide clear explanations for their predictions or decisions.
Explainable AI (XAI): A research area in artificial intelligence that aims to develop AI systems that can provide explanations or justifications for their decisions and actions, in order to enhance transparency, accountability, and trustworthiness.
Fairness: The property of a machine learning model that ensures that the predictions or decisions it makes do not discriminate against particular groups of people or result in unjust outcomes.
Feature Engineering: The process of selecting and transforming the most relevant features of a dataset to improve the accuracy of a machine learning model.
Federated Learning: A type of machine learning that allows multiple parties to collaboratively train a model without sharing their data, by using local models to compute gradients and aggregating them at a central server.
Few-Shot Learning: A type of machine learning where the goal is to learn from a small number of examples, rather than a large dataset.
Friendly AI: A concept in artificial intelligence safety that refers to an AI that is aligned with human values and goals, and does not pose a threat to humanity or the environment.
Gated Recurrent Unit (GRU): A type of recurrent, neural network architecture that uses gating mechanisms to selectively update and forget information in the hidden state, making it more efficient than traditional recurrent neural networks.
Generative Adversarial Network (GAN): A type of neural network used in unsupervised learning to generate new data that is similar to the training data.
Generative Adversarial Networks (GANs): A type of neural network architecture that consists of a generator network that learns to generate samples from a target distribution, and a discriminator network that learns to distinguish between the generated samples and real samples from the target distribution.
Gradient Boosting: A machine learning technique that combines multiple weak models to,improve accuracy and reduce error.
Gradient Descent: A technique used in machine learning to optimize a model by adjusting its parameters to minimize the error.
Human-in-the-Loop (HITL): A machine learning workflow that involves humans in the training or validation process, to provide feedback and ensure the model is performing as desired.
Hyperparameter: A parameter in a machine learning model that is set prior to training, such as learning rate or number of hidden layers.
Kernel Trick: A technique used in machine learning to transform a dataset into a higher-dimensional space to make it easier to separate and classify.
Large language model (LLM): Computer program for natural language processing that uses deep learning and neural networks.
Long Short-Term Memory: A type of recurrent neural network that is designed to remember information over long periods of time.
Long Short-Term Memory (LSTM): A type of recurrent neural network (RNN) that is capable of remembering long-term dependencies.
Long Short-Term Memory (LSTM): A type of recurrent neural network architecture that uses memory cells and gating mechanisms to selectively update and forget information in the hidden state, making it well-suited for tasks involving long-term dependencies.
Machine Learning: A subset of AI that allows computers to learn and improve on their own without being explicitly programmed.
Machine Learning Operations (MLOps): The process of managing and automating machine learning workflows, including data management, model training, and deployment.
Markov Decision Process (MDP): A formalism for modeling decision-making problems, where an agent interacts with an environment over time, receiving rewards or punishments based on its actions, and tries to learn a policy that maximizes the expected cumulative reward.
Meta-Learning: A type of machine learning where the goal is to learn how to learn, by training models that can adapt quickly to new tasks or environments.
Meta-Learning: A type of machine learning that learns to learn, by discovering and adapting to the structure of a set of tasks or environments, in order to improve its generalization and transferability.
Model Selection: The process of selecting the best model for a particular problem based on its accuracy and generalization performance.
Model-Based Reinforcement Learning: A type of reinforcement learning that learns a model of the environment, such as a dynamics model or a transition function, and uses it to plan and optimize actions, in order to improve sample efficiency and robustness.
Model-Free Reinforcement Learning: A type of reinforcement learning that learns to map states to actions directly, without explicitly modeling the environment, by using trial-and-error to explore and exploit the reward signal.
Monte Carlo Tree Search (MCTS): A search algorithm that uses a tree structure to represent the possible actions and states in a game or decision-making problem, and simulates many possible trajectories to estimate the expected reward of each action.
Named Entity Recognition (NER): A type of NLP that involves identifying and extracting named entities such as people, places, and organizations from text.
Natural Language Processing: The ability of computers to understand, interpret, and generate human language.
Natural Language Processing (NLP): A field of study that combines computer science and linguistics, focused on developing algorithms to analyze and understand human language.
Neural Network: A network of interconnected nodes that are designed to simulate the function of a human brain and can be used for tasks such as image and speech recognition.
Neural Network: A type of machine learning model that is inspired by the structure and function of the human brain, consisting of multiple interconnected layers of neurons.
Neural Style Transfer: A technique that uses a neural network to transfer the style of one image onto another image, creating a hybrid image that combines the content of one image with the style of another.
Neuroevolution: A type of machine learning that uses evolutionary algorithms, such as genetic algorithms or particle swarm optimization, to evolve neural networks that can perform a given task, by selecting and mutating individuals based on their fitness.
Off-Policy Reinforcement Learning: A type of reinforcement learning that learns from a different policy than the one used to generate the data, by using importance sampling or other techniques to estimate the value of actions or states under a different policy.
On-Policy Reinforcement Learning: A type of reinforcement learning that learns from the same policy that generates the data, by using the trajectory or episode data to update the policy directly, such as with policy gradient or actor-critic methods.
One-Shot Learning: A type of machine learning where the goal is to learn from a single example, rather than a large dataset.
One-Shot Learning: A type of machine learning that can learn from a single or a few examples of a new class or task, by using transfer learning, generative models, or metric learning, to recognize or synthesize new instances from the limited data.
Overfitting: A common problem in machine learning where a model is too complex and fits the training data too well, resulting in poor performance on new data.
Partially Observable Markov Decision Process (POMDP): A variation of the Markov decision process where the agent only partially observes the environment, and must infer the underlying state from the observations.
Policy Gradient: A type of reinforcement learning algorithm that learns to directly optimize a policy for selecting actions, by using gradient ascent on the expected reward of the policy.
Principal Component Analysis (PCA): A technique used in unsupervised learning to reduce the dimensionality of a dataset by finding the most important features or components.
Privacy-Preserving Machine Learning: The practice of developing machine learning models that preserve the privacy of sensitive data, by using techniques such as differential privacy or secure multiparty computation.
Q-Learning: A type of reinforcement learning algorithm that uses a table to store the expected future rewards for each state-action pair, and learns to select actions that maximize the expected future reward.
Q-Learning: A type of reinforcement learning that learns a state-action value function, called Q-function, by using a temporal difference algorithm and an exploration strategy, such as epsilon-greedy or softmax, to balance between exploration and exploitation.
Recurrent Neural Network: A type of neural network that is designed to process and analyze sequential data such as text and speech.
Recurrent Neural Network (RNN): A type of neural network that is capable of processing sequential data by maintaining an internal state or memory of previous inputs.
Regression: A technique used in supervised learning to predict a continuous output variable based on input variables.
Reinforcement Learning: A type of machine learning in which an algorithm learns through trial and error by receiving feedback in the form of rewards or punishments.
Reinforcement Learning: A type of machine learning that involves an agent learning from trial and error by receiving rewards or penalties based on its actions.
Reinforcement Learning: A type of machine learning where an agent learns to make decisions by interacting with an environment, receiving rewards or punishments based on its actions.
Reinforcement Learning: A type of machine learning that involves an agent interacting with an environment, in order to learn a policy or a value function that maximizes a reward signal, by using trial-and-error and exploration-exploitation strategies.
Restricted, Boltzmann Machine (RBM): A type of generative model that consists of a bipartite graph of visible and hidden units, with binary stochastic activations, and learns to represent the input data distribution by adjusting the weights between the units.
Robotics: The branch of AI that deals with the design and construction of robots that can perform tasks autonomously.
Sentiment Analysis: A type of NLP that involves analyzing the emotions and opinions expressed in text or speech.
Sim-to-Real Transfer: A research area in robotics that aims to transfer control policies or models trained in simulation to the real world, by adapting to the differences in dynamics, sensors, and perception, in order to reduce the cost and risk of real-world experiments.
Singularity: A hypothetical event in the future when artificial intelligence surpasses human intelligence, leading to a dramatic transformation of society and possibly the extinction of humanity.
Superintelligence: A hypothetical level of artificial intelligence that surpasses human intelligence in every cognitive task, and has the ability to improve itself recursively, leading to an intelligence explosion and potentially catastrophic consequences.
Supervised Classification: A type of supervised learning used for classifying data into pre-defined categories.
Supervised Learning: A type of machine learning in which an algorithm is trained on labeled data to make predictions on new data.
Transfer Learning: A technique in machine learning that allows a model to reuse knowledge gained from one problem to improve performance on a new, related problem.
Transfer Learning: A type of machine learning where a model is first trained on one task, and then fine-tuned on a different but related task.
Underfitting: A common problem in machine learning where a model is too simple and fails to capture the complexity of the data, resulting in poor performance on both training and new data.
Unfriendly AI: A concept in artificial intelligence safety that refers to an AI that is misaligned with human values and goals, and could pose a threat to humanity or the environment if it gains too much power or autonomy.
Unsupervised Learning: A type of machine learning in which an algorithm is trained on unlabeled data to discover patterns and relationships on its own.
Unsupervised Learning: A type of machine learning where the algorithm learns patterns and relationships in the data without explicit supervision or labels.
Value Alignment: A research area in artificial intelligence safety that aims to ensure that advanced AI systems are aligned with human values and goals, and do not cause unintended harm or adverse consequences.
Variance: The amount by which a machine learning model's predictions vary for different training datasets.
Variational Autoencoder (VAE): A type of neural network architecture that consists of an encoder network that learns to encode input data into a low-dimensional representation, and a decoder network that learns to reconstruct the input data from the representation, while imposing a regularization constraint on the representation to encourage diversity.
Zero-Shot Learning: A type of machine learning that can generalize to unseen classes or tasks, by using prior knowledge, such as attributes or semantic embeddings, to infer the properties of the new classes or tasks, without requiring explicit training examples.