Concepts starting with G

99 concepts, each with a plain-language explanation, examples and diagrams. Pick a letter or open a concept.

Game-Playing Programs and Lookahead Search — Heuristic search decides how much of the possible game tree to examine.
GELU Activation Function — ReLU outputs its input for positive values and zero for any negative input, which creates a sharp corner at zero.
General Books on Reinforcement Learning — Use the purpose of your reading to choose among the listed resource categories.
Generalization — Overfitting occurs when a hypothesis fits the training data too well but performs poorly on the true distribution.
Generalization and Function Approximation in Reinforcement Learning — A policy can be represented by a parameterized function whose parameters form the weight vector θ.
Generalization and Learnability — ϵ-representativeness describes a condition on a training set S relative to Z, H, ℓ, and D.
Generalization and Learning Guarantees — The No-Free-Lunch Theorem rules out a learner that succeeds on every learning task.
Generalization Bounds — Margin separation supplies the condition for the Perceptron convergence guarantee.
Generalization Bounds for Linear Predictors — Hard-SVM's output w_S is defined by Equation (26.19) and satisfies L_S(w_S) = 0 under the theorem's assumptions.
Generalization Bounds for Predictors with Low ℓ1 Norm — Theorem 26.12 and Theorem 26.15 have bounds that look similar apart from an extra log(d) factor in Theorem 26.15.
Generalization from Training Data — ERM can overfit because success on a training sample does not ensure success over the underlying data distribution.
Generalization in Linear Function Approximation — Receptive-field size and shape determine the initial generalization behavior of a linear function approximation method.
Generalization in Machine Learning — Shattering means realizing every possible 0/1 labeling on a chosen set.
Generalization in Tiling Strategies — Multiple tilings provide the overlapping receptive fields needed for coarse coding.
Generalization Risk — i.i.d. sampling means independent examples drawn from the same distribution D.
Generalized Policy Iteration — Evaluation and improvement impose different requirements on a policy-value pair.
Generalized Policy Iteration: A Unified Framework — GPI converges through the interaction of policy evaluation and policy improvement.
Generalizing Customer-Specific Features — A Forward Deployed Engineer builds features for a specific customer while working in the field.
Generalizing the Learning Model — True error evaluates a prediction rule h with respect to labeled points drawn according to the probability distribution D.
Generating Random Floating-Point Numbers with random.random() — random.randint(low, high) generates a random integer within an inclusive range, including both the low and high endpoints.
Generating Random Integers and Choices — Continuous distributions generate random values from an infinite range, unlike discrete selection which picks from a fixed set.
Generating Random Numbers in Java — The exercise splits the program into main(), which repeats games, and the static subroutine playGame(), which plays one round.
Generating Text with a Simple Loop — The modified generate function adds top-k filtering and temperature scaling to the earlier loop.
Generating text with generate_text_simple — generate_and_print_sample takes a starting text, encodes it into token IDs, and asks the LLM to generate a continuation of up to…
Generating Text with GPT — text_to_token_ids uses a tokenizer's encode method to turn text into token IDs, and then adds a batch dimension with unsqueeze(0).
Generating Text with the generate Function — Loading pretrained GPT-2 weights means calling load_weights_into_gpt with the gpt instance and the params, then moving the model…
Generative Adversarial Networks — A latent space of images is a low-dimensional vector space whose points can be mapped to realistic-looking images.
Generative Deep Learning — Variational autoencoders are generative models that learn compressed latent representations.
Generative Models — Chapter 5 examines practical computer-vision examples with a focus on image classification.
Generator Expressions — List comprehensions are a concise syntax for creating a new list by transforming or filtering elements from an existing list.
Generator Expressions and Lazy Evaluation — List comprehension condenses a for loop, empty list initialization, and append pattern into a single, readable line.
Generic Classes and Formal Type Parameters — Wildcard types let a method work with collections of objects of various types, but they do not restrict the types allowed as…
Generic Classes and Methods — A type parameter that can be any type limits you to operations that every type supports.
Generic Classes in Java — Generic programming is a paradigm that produces highly general and reusable code, meaning code not tied to one narrow situation.
Generic Classes with One Type Parameter — A generic class with two type parameters declares both names in one angle-bracket list after the class name, separated by a…
Generic Method Parameters — A collection of a subclass, such as Collection<Rect>, is a different type from a collection of its superclass…
Generic Methods — A type parameter is a placeholder written in angle brackets after a class name, and it works much like a formal parameter in a…
Generic Programming in Other Languages — In Smalltalk, variables are typeless and values carry the types, so one data structure or subroutine can serve any type of data.
Generic Programming with Collection<T> — An iterator is an object that traverses a collection, and the same calling pattern works across different collection types.
Generic Types and Type Parameters — Function<T,R> represents functions from values of type T to values of type R, and it defines apply(t) where t is of type T and…
Geometric Classifiers — Shattering is an all-labelings requirement, not a one-labeling test.
Getting Input from Users: The input() Function — ValueError is a runtime error that occurs when type conversion fails—for example, when int() tries to convert a string like…
Getting Started with Deep Learning — Keras is the open-source, free-to-download framework used by the course examples.
Getting Started with Keras — The main practical prerequisite is reasonable Python proficiency for using Keras.
Getting started with neural networks — Deep learning became more popular through the combined influence of hardware, data, algorithms, investment, and democratization.
Getting Started with the Python Interpreter — Python's built-in help() function provides immediate access to documentation about functions, statements, and objects without…
Getting Started: Why Programming Matters and How Computers Work — Programming languages, like natural languages, have vocabulary (words that must be spelled correctly) and grammar (rules for…
Getting User Input with input() — try-except blocks allow programs to catch and handle errors gracefully instead of crashing when unexpected situations occur.
Global and Local Variables — A subroutine can use local variables, formal parameter names, and static member variables declared outside the subroutine.
Global Variables and the Global Keyword — If you want to assign a value to a name defined at the top level of the program (i.e. not inside any kind of scope such as…
Go-to-Market Strategy for Specialized Customers — The company was founded in October 2003 by a team of five co-founders, and its first product was built for the intelligence…
Goal Gradients and Delayed Reinforcement — A delayed primary reward does not have to be the only influence operating during the delay.
Goal-Directed Learning — Reinforcement learning combines learning with decision-making in a computational setting.
Goals and Uncertainty in Artificial Intelligence — The reinforcement learning framework describes interaction between a learning agent and an environment.
GPT Model Architecture and Its Configuration — BASE_CONFIG holds the shared GPT-2 settings, and BASE_CONFIG.update copies the chosen size's emb_dim, n_layers, and n_heads into…
GPT Model Sizes — A transformer block is a core structural component of GPT models that combines masked multi-head attention modules with fully…
Gradient Ascent for Policy Optimization — The policy gradient theorem links performance changes to the policy-weight vector θ.
Gradient Bandit Algorithms for Action Selection — The gradient bandit algorithm can be understood as a stochastic approximation to gradient ascent.
Gradient Descent — Early neural networks were explored decades before modern deep learning, but their development was constrained by training…
Gradient Descent Algorithm — The gradient identifies the direction in which the function increases fastest around the current vector.
Gradient Descent and Backpropagation — Part 1 is a staged introduction rather than a single isolated topic.
Gradient Descent and Loss Functions — VGG19 acts as a pretrained measurement network rather than as the object being generated.
Gradient Descent and Stochastic Gradient Descent — A gradient at w produces the first-order Taylor approximation of f around w.
Gradient Descent Bound — SGD requires an expectation-based analysis.
Gradient Descent for Value-Function Approximation — The Gradient Monte Carlo Algorithm evaluates a policy by adjusting the weights of a differentiable value-function approximation.
Gradient MC — SGD variations are used to find a good weight vector for function approximation.
Gradient Monte Carlo Methods — Monte Carlo updates depend on the final return, so the weight vector is updated only after an episode ends.
Gradient Monte Carlo Prediction — Monte Carlo updating is delayed because the final return is needed.
Gradient of a Function — The gradient identifies the direction in which the function increases fastest around the current vector.
Gradient Propagation — Data provides the material from which deep-learning systems learn.
Gradient-Based Optimization — Tensors provide the multi-dimensional representations used by neural networks.
Gradients — A subgradient is a vector that supplies a lower-supporting linear slope at a point.
Gradients and Directional Change — A partial derivative measures the rate at which a multivariate function changes with respect to one of its variables.
Gradients and Weight Updates in Neural Networks — Training deep neural networks with many layers can be challenging, and the source names vanishing or exploding gradients as a…
Graduate Course Structure — The supplied material describes a 14-week graduate introduction course.
Graph Coloring and Reductions — NP-hardness can arise in ERM implementation even for a network with one hidden layer and only four hidden neurons.
Graph-based clustering — A similarity graph represents each data point as a vertex and each pairwise similarity as an edge weight.
Graphics and Animation Examples — GrowingCircleAnimation.java, from Section 5.3 of the text, shows an animation of growing, semitransparent circles.
Graphics Class Basics in Java — All drawing in Java is done through a Graphics object, which provides basic commands for shapes, text, and color.
Graphics2D Drawing Basics — Aliasing is the jagged appearance of shapes drawn on a pixel grid, and antialiasing is the technique that reduces that…
Greedy and Epsilon-Greedy Action Selection — Heuristic search looks ahead from possible actions and uses backed-up rewards and estimated values to guide a choice.
Greedy and Exploratory Action Selection — High initial action-value estimates can make a greedy method explore without explicitly using ε-greedy random selection.
Greedy and Exploratory Policies — Off-policy learning separates the policy being evaluated from the policy generating experience.
Greedy and Lazy Quantifiers: Controlling How Much to Match — The * quantifier matches zero or more characters, while + matches one or more characters.
Greedy and Nongreedy Actions — Exploitation uses the action that currently has the greatest estimated value.
Greedy and ε-Greedy Action Selection — The 10-armed testbed uses 2000 different ten-action bandit problems to compare learning behavior under averaged results.
Greedy Policy — Policy iteration alternates between evaluating a policy and improving it.
Greedy vs. Non-Greedy Matching — The $ anchor marks the end of a line in regular expressions, not a literal dollar sign character
Grid Layout in Java Swing — A grid layout can only have one component in each position, so a row that needs two pieces cannot place both directly into one…
Gridworld Examples 13.1 and 13.2 — This practice section applies REINFORCE to gridworlds.
Gridworld Value Functions — v*(s) is the maximum expected return from a particular state under an optimal policy.
Grouping and Capturing with Parentheses — re.search() returns a match object when a pattern is found, or None when no match exists.
Grouping Radio Buttons with ButtonGroup — A JCheckBox draws its box from a pair of Icons: one for the unselected state and one for the selected state.
Growing Customer Relationships through AI Expansion — In successful AI startups, contract size tends to grow over the life of a customer relationship.
Growth Functions — A class with small effective size enjoys the uniform convergence property.
Growth Mindset and Persistence in Programming — Programming is a language that requires gradual absorption over time, not linear mastery of isolated concepts.
GUI Design Patterns and Best Practices — Python bindings are translation layers that allow Python code to use GUI libraries written in C/C++, combining Python's…
Guido van Rossum and Python's Design Philosophy — Guido van Rossum, the creator of the Python language, named the language after the BBC show "Monty Python's Flying Circus". He…
Guido van Rossum and Python's Design Philosophy — Python was created by Guido van Rossum, who named the language after the British comedy show 'Monty Python's Flying Circus'…