Atlas
Concepts.
The connected concept graph behind every course.
Explore Python concepts from the basics to advanced topics. Each concept is explained in simple language with examples, diagrams and real-world use cases.
- 5485 Concepts
- 12 Core Topics
- A–Z glossary
- Visually explained
Core Topics
Browse all 54855485 concepts
#7 concepts
0-1 LossThe multiclass categorization goal is to learn h : X → [k].19-state random walk taskThe λ-return is presented as a smooth alternative between Monte Carlo and one-step TD met…1D ConvnetsSequence models operate on ordered numeric representations, not raw text.ε-Greedy Action SelectionHigh initial action-value estimates can make a greedy method explore without explicitly u…ε-Greedy PoliciesSarsa control estimates the action values of its current behavior policy.ε-soft policiesOn-policy first-visit MC control alternates between episode generation and policy improve…λ-Return Update TargetsThe h-truncated λ-return method expands its available information one horizon at a time.
A273 concepts
A Clustering ModelThe supplied source pack does not define PCA or variance maximization.A First Look at a Neural NetworkAdvanced deep learning combines tensor-based data representation, tensor operations, and…A Trivial Client/Server ExampleDateServer creates a ServerSocket on LISTENING_PORT, announces that it is listening, and…A Year in the Life of a Forward Deployed EngineerOnboarding for a Forward Deployed Engineer combines meetings with senior engineers, readi…A/B Testing for AI ModelsComparative evaluation selects two or more models to respond to the same request, and an…Absolute Positional EmbeddingsA token embedding is a vector representation of an input token, created in PyTorch with a…Abstract Base Classes and InterfacesPolymorphism allows a subtype to be treated as an instance of its parent type, enabling f…Abstract ClassesAn abstract class is declared with the abstract modifier and is meant to be subclassed ra…
B166 concepts
Background Computation with ThreadsA shared resource is an object that more than one thread uses, such as the image object i…BackpropagationTensors provide the multi-dimensional representations used by neural networks.Backpropagation and the computational graphtorch.no_grad() is a context manager that tells PyTorch not to track gradients, which sav…Backpropagation and Weight UpdatesIdentical starting weights make nodes in a layer perform the same operations and receive…Backpropagation in Neural Network LayersThe embedding layer is a lookup operation: a token ID selects a row of the weight matrix…Backward EliminationGreedy selection evaluates feature subsets through the learning algorithm rather than sep…Backward-view temporal-difference learningThe on-line λ-return algorithm motivates this topic because it combines strong performanc…Balancing and Preprocessing the DatasetUndersampling balances a dataset by reducing the majority class until it matches the mino…
C499 concepts
C Identifiers and KeywordsC++ has many more keywords than C, and any C++ keyword used as an identifier in C code mu…C-Style String RepresentationPositions in a string run over a fixed range, and a position outside that range does not…C++ Class Design Best PracticesA concrete type is the simplest kind of class.C++ Evolution and StandardizationA language standard does not reliably finish on the schedule its early planning suggests;…C++ Function DefinitionsA function's interface names the function, its return type, and its parameters.C++ Implementations and Their FacilitiesStandard-library components are part of every complete C++ implementation and are portabl…C++ Mathematical Functions OverviewThe mathematical functions for complex numbers are provided in the <complex> module.C++ Standards and the Move to C++11C and C++ are separate languages with their own standards, so code valid in one is not au…
D425 concepts
Data Access in AI EngineeringA cargo cult engagement has the title and vocabulary of a successful AI engagement but la…Data Acquisition and AnnotationDiminishing returns are typical: each additional block of examples usually yields a small…Data Augmentation and SynthesisAnnotation is hard because of the process and because creating clear guidelines is comple…Data Augmentation for AI TrainingProcedural generation means using algorithms to generate data, as opposed to generating i…Data Challenges in AI DeploymentsProduction data should be examined early in a project, not late.Data Cleaning and Deduplication for TrainingData for finetuning must be in the format the target model expects, which includes the to…Data Cleaning and Normalization in Large ArchivesMapping tables unify changing email addresses and multiple domain names into single canon…Data CompressionUnsupervised learning finds useful transformations of input data without target values.
E270 concepts
Earlier NLP Models and Task-Specific DesignLLMs are AI models that can understand, generate, and interpret human language, where und…Echo and Delta: Two Halves of One CraftFounder-phase skills are the same skills the FDE role requires, and the labor market valu…Echo Mode: Listening Before BuildingEcho mode listens and learns how the work actually happens; Delta mode ships a working sy…Economic Viability of Go-to-Market MotionsThe viability of the FDE motion depends on a company's average contract size, not on the…Effect of Trainable Parameters on MemoryWeight memory equals the parameter count multiplied by the bytes per parameter; a 13B mod…Effective Communication with Foundation ModelsA language model, by itself, does not differentiate between user-provided input and its o…Effective Contract Negotiation for AI ProjectsA contract's structure, especially how payments are scheduled and released, determines wh…Effective Data Preparation for AI ModelsInspecting a raw dataset comes before training and aims to judge its quality and to revea…
F213 concepts
Failure Mode Analysis for AI SystemsA validation gate is a checkpoint that measures how a system performs before it is deploy…Fall Through and Case GroupingThe traditional switch uses a colon after each case label, and a case label may hold one…FDE Work from Chapter 12Production engineering instinct is the understanding of a system's operational state: kno…Feature ConstructionRBF networks are function approximators whose features are radial basis functions.Feature Construction for Linear Function ApproximationEach OrderN polynomial basis function multiplies one powered term for every state variabl…Feature Delivery Engineering in AI ProductsAn AI engineering role can be self-terminating for a single product: when the engineer do…Feature EngineeringA machine learning problem is defined by its inputs, predicted outputs, and problem type.Feature extractionPreprocessing changes raw data into numerical forms that neural networks can use.
G115 concepts
Game-Playing Programs and Lookahead SearchHeuristic search decides how much of the possible game tree to examine.GELU Activation FunctionReLU outputs its input for positive values and zero for any negative input, which creates…General Books on Reinforcement LearningUse the purpose of your reading to choose among the listed resource categories.GeneralizationOverfitting occurs when a hypothesis fits the training data too well but performs poorly…Generalization and Function Approximation in Reinforcement LearningA policy can be represented by a parameterized function whose parameters form the weight…Generalization and Learnabilityϵ-representativeness describes a condition on a training set S relative to Z, H, ℓ, and D.Generalization and Learning GuaranteesThe No-Free-Lunch Theorem rules out a learner that succeeds on every learning task.Generalization BoundsMargin separation supplies the condition for the Perceptron convergence guarantee.
H129 concepts
Habit Learning and Goal-Directed BehaviorGoal-directed and habitual behavior rely on different, though not necessarily exclusive,…Habitual and Goal-Directed Processes in the BrainModel-free and model-based reinforcement learning provide a distinction for studying brai…Habitual BehaviorThe experiment changes reward value after learning and measures whether behavior adjusts.Halfspace HypothesesThe same dot product ⟨w, x⟩ can be interpreted probabilistically by logistic regression o…Halfspace LearningSurrogate losses replace other losses for purposes such as computational convenience.HalfspacesThe base class B provides the individual hypotheses used as building blocks.Hallucination in Language ModelsTemperature controls how random a model's output is. Higher values let lower-probability…Hand-Calculating Layer ConnectionsThe source's network has two inputs, two outputs, and two hidden layers of 30 and 20 node…
I346 concepts
ID3 Decision Tree LearningDecision tree learning searches for a tree that minimizes the relevant bound, but solving…Identifiers and VariablesA variable names a memory location rather than being the data itself.Identifying Cargo Cult EngagementsThe cargo-cult engagement diagnostic is a tool for identifying cargo-cult engagements by…Identifying Repeated Work Across EngagementsA pattern becomes a product feature through four stages: recurring engagements, a selecte…Identifying the Real ProblemA stated problem is how a person describes a problem, while the real problem is the genui…Identity and Access in Agent SystemsA shared service account can compromise the compliance posture for material non-public in…Identity in System DesignA complete audit log is not sufficient for compliance if its entries cannot identify the…If StatementsOrdinary program flow executes instructions in their sequence.
J46 concepts
Jacobian MatrixBackpropagation calculates the gradient of a loss with respect to network weights for an…Jar Files and Packaging ProgramsA resource is data a program uses that is not code, such as images, sounds, and text; the…Java and Networked ApplicationsThe Internet is a worldwide network containing millions of connected computers.Java and the InternetJava's networking support should be understood as one advantage among others.Java Arrays BasicsThe standard array-processing methods are static methods defined in the Arrays class, whi…Java bytecodeJust-in-time compilation translates Java bytecode into native machine language during pro…Java Classes and Instance VariablesA Java record is an object that belongs to a special kind of class called a record class.Java Classes and InterfacesPackages are named containers for related Java classes and interfaces.
K19 concepts
k Nearest Neighbors AlgorithmThe k-NN rule maps a training sample and a chosen k to a label for each point under consi…k-Means and Other Cost Minimization ClusteringsThe objective function and the iterative algorithm are related but not identical ideas.k-Means ClusteringDictionary learning creates a feature vocabulary for data that may not have an obvious vo…K-Nearest Neighbors for RetrievalHybrid search combines term-based retrieval with embedding-based retrieval, and the two c…k-NN Classificationk-NN regression predicts a real-valued target by combining the targets of the k nearest n…Kamin BlockingThe Rescorla-Wagner model links learning to surprise.KerasThe convnet progressively exchanges spatial detail for deeper feature representations.Keras and Deep Learning FrameworkPart 1 is a staged introduction rather than a single isolated topic.
L223 concepts
Labeling Cases for EvaluationAn evaluation set is a collection of contracts with known correct answers, and it can gro…Labeling Interfaces for Evaluation WorkAn evaluation set should be co-created with the customer, not written by the development…Labeling Pipelines in ValidationValidation is a separate phase from delivery. The source states that validation is not de…Lambda ExpressionsA functional interface is a Java interface with a single abstract method, and Runnable is…Lambda Expressions and Variable AccessA local variable is effectively final if its value is never changed after it is initializ…Lambda Expressions as Arguments to AlgorithmsA lambda expression is a notation that generates a function object, similar in kind to wr…Lambda Expressions in AlgorithmsA lambda expression is a notation that generates a function object, exactly like Less_tha…Lambda Expressions in JavaAn ActionListener is a functional interface whose single method, actionPerformed, receive…
M256 concepts
Machine Code and Binary: What the CPU Actually UnderstandsThe CPU is the part of the computer built to obsessively ask 'What is next?' and determin…Machine LanguageA CPU directly executes machine language, not a high-level program.Machine Language and Binary RepresentationA compiler translates an entire program from high-level source code to machine language i…Machine Language Instruction ExecutionThe CPU's normal behavior is a sequential fetch-and-execute process.Machine Language InstructionsA program is stored as a sequence of machine language instructions in main memory.Machine LearningDeep learning is a subset of machine learning.Machine Learning Data TerminologyA class is a category in a classification problem.Machine Learning DatasetsData provides the material from which deep-learning systems learn.
N126 concepts
n-step action-value backupsn-step Q(σ) unifies Sarsa and tree backup by allowing sampling and expectation to be sele…n-step Action-Value Backups with Q(σ)ε-greedy behavior is defined relative to the current Q-values of the target policy.n-Step BackupsThe full return is the complete discounted reward sequence through episode termination.n-Step BootstrappingThe differential n-step return is G(n)t = R̄ + V̄(S(t+n)) − V̄(S(t)).n-step Expected Sarsan-step Q(σ) is a unifying framework for action-value backups.n-step MethodsOff-policy learning separates the policy collecting experience from the policy being lear…n-Step Methods for Reinforcement Learningn-step semi-gradient Sarsa joins a semi-gradient Sarsa control method with n-step bootstr…n-step Returns and the λ-returnThe forward view of eligibility traces connects eligibility-trace ideas with n-step retur…
O164 concepts
OAuth 2.0 Authorization Code FlowOAuth libraries are free, pre-built tools that handle the complexity of the OAuth 2.0 pro…OAuth and Token-Based AuthenticationAPI keys are the mechanism vendors use to identify users and monitor their service consum…Object Attributes and MethodsObjects have a complete lifecycle: creation (via __init__), active use, and destruction (…Object Construction and InitializationAn object's lifetime runs from its construction, which must finish before use, to its des…Object Creation in C++The read_and_sum() function is a client of the Vector class: it creates a Vector, reads a…Object Equality and the equals() MethodA record declaration with only fields causes the compiler to add a canonical constructor,…Object Initialization and CleanupSpecial methods enable custom classes to mimic built-in type behaviors through double-und…Object Initialization and ConstructorsA constructor has the same name as its class, has no return type, and initializes an obje…
P318 concepts
PAC LearnabilityShattering means realizing every possible 0/1 labeling on a chosen set.PAC LearningRealizability demands perfect agreement between some hypothesis in H and a target labelin…PAC Learning and Sample ComplexityUniform convergence connects a finite i.i.d. sample with representativeness of an underly…PAC Learning ModelOnline learning has no separate training phase followed by a separate prediction phase.PAC-Bayes BoundsThe PAC-Bayes bound can be converted into a learning rule.PAC-Bayes Bounds and Measure ConcentrationPAC-Bayes bounds define a hierarchy over a hypothesis class H.Pacing Delta Mode Against Institutional RealityEcho mode listens and learns how the work actually happens; Delta mode ships a working sy…Package Statements in Java Source FilesAn import directive lets you write a simple class name in the rest of a Java file instead…
Q26 concepts
Q-learning: Off-policy TD ControlThe cliff-walking task separates the shortest optimal route from the safer route under ex…Q-value estimationOff-policy n-step Q(σ) separates the policy that generates experience from the policy who…Quality Control for Training DataDataset engineering covers curating, generating, and annotating the data used to train an…Quality failures in model outputsAn output guardrail catches failures in model output and specifies a policy for each fail…Quantifiers and Repetition PatternsThe backslash (\) is the escape character that converts special regex characters into lit…Quantifiers: Matching Multiple CharactersThe period (.) in a regular expression is a wildcard that matches any single character—le…Queries, Keys, and ValuesThe SelfAttention_v1 class stores three trainable matrices, W_query, W_key, and W_value,…Query Caching CriteriaExact caching reuses a stored result only when the exact same item is requested, so the l…
R267 concepts
Race Conditions in Shared DataA condition variable is a mechanism that allows one thread to wait for another thread.Rademacher ComplexityThe chaining technique uses covering numbers to bound Rademacher complexity.Rademacher Complexity and Generalization BoundsRademacher complexity is a measure of the complexity of a set of functions or vectors.Rademacher Complexity for Linear PredictorsHard-SVM's output w_S is defined by Equation (26.19) and satisfies L_S(w_S) = 0 under the…Rademacher Complexity of Linear ClassesThe H1 proof reduces a Rademacher-complexity calculation to a finite-set problem.RAII and Scope-Based Resource ManagementA resource handle is an object that manages a resource, and it is distinct from the resou…Raising and catching custom exceptionsThe else block in exception handling runs only when the try block succeeds without raisin…Raising and Handling ExceptionsCustom exceptions are user-defined classes that inherit from Exception, enabling you to r…
S471 concepts
Sample ComplexityRegression loss functions measure more than whether a prediction is wrong: they quantify…Sample Complexity of Finite Hypothesis ClassesThree conjunctions connected by OR define a 3-term DNF.Sample Complexity of Neural NetworksVC dimension is a measure of hypothesis-class capacity based on shattering.Sample ModelPlanning uses model-generated experience, while learning uses experience from the real en…Sample Models in BlackjackThe key distinction is whether the model exposes the full probability description or only…Sample Size FlexibilityThe two notions share the same competitiveness requirement.Sample-Average Action-Value EstimationA sample average is a natural estimate of an action value.Sample-Average Action-Value MethodsInitial estimates affect action-value methods because they provide the starting values us…
T420 concepts
Tabular n-step TD algorithmThe semi-gradient n-step TD algorithm is the function-approximation extension of tabular…Tabular Solution Methods in Reinforcement LearningBandit problems are the single-state special case of reinforcement learning.Tabular TD(0) for Estimating Value FunctionsThe TD error is the difference between the current estimate V(S_t) and the one-step estim…Tag Dispatch and Iterator CategoriesThe standard header <iterator> provides the class template iterator_traits, which supplie…Task Decomposition and Chain-of-Thought PromptingChain-of-thought (CoT) prompting explicitly asks a model to think step by step, nudging i…Task Division in Parallel ProcessingEqual-sized subproblems do not guarantee equal completion times, because the work inside…Task-Based Parallelism with packaged_taskTask-based parallelism executes tasks concurrently, and packaged_task is a template used…TD Control MethodsThe cliff-walking task requires safe movement from S to a goal in a grid world.
U371 concepts
Unary and Binary OperatorsAn overloaded arithmetic operator is a function whose name is the operator symbol, such a…Uncertainty in EnvironmentGoal-directed agents pursue explicit goals through sensing and action.Under-representation of languages in training dataUnder-representation in training data is a major reason models perform worse in some lang…Understand Boolean ExpressionsAlternative execution uses if-else to run one of two code blocks based on whether a condi…Understand Comparison OperatorsAlternative execution uses if-else to run one of two code blocks based on whether a condi…Understanding __init__ and the Constructorsuper() is a special syntax that calls a method from the parent class. In the context of…Understanding Action PropertiesAn Action can be passed to the constructor of a JButton or a JMenuItem, so one Action can…Understanding Actual Workflows through Discovery InterviewsA discovery interview, as the source presents it, is a conversation of about thirty minut…
V96 concepts
Validating AI Systems Against Real DataProduction data should be examined early in a project, not late.ValidationApproximation error measures the gap between the best hypothesis in a class and the true…Validation and Model EvaluationThe usefulness of a learnability definition depends on the goal for which it is being use…Validation Gates in Production SystemsThe pilot was staged, starting with three inspectors at one plant and growing to twenty a…Validation Loops and Feedback SignalsAgent readiness is a measure of how many deterministic validation loops, such as linters,…Validation TechniquesA model-selection curve compares training and validation error across model complexity.Value BackupsBackward focusing directs planning from a changed value toward states that may depend on…Value Estimates and Backup MethodsHeuristic search focuses on the current state and its likely successors.
W227 concepts
Waiting for Threads with joinA thread can be given a function and its arguments, and the function then runs in a separ…Watson's Daily-Double Wagering StrategyWatson evaluated wagers as actions rather than judging them only by their point amounts.Watson's Question Answering and Game-Playing ArchitectureWatson treated Daily-Double wagering as a decision under uncertainty rather than as a fix…Weak LearnabilityA 3-piece classifier operates on the real line and is specified by two real thresholds th…Weak LearnersBoosting provides a way to manage the tradeoff between approximation error and estimation…Web Crawling and Data Collection with spider.pyPageRank is computed by running sprank.py, which iteratively calculates importance scores…Web Crawling with spider.pyVisualizing a network graph requires running spjson.py to convert your PageRank database…Web Scraping Basicscurl and wget are command-line tools for retrieving files from the web on Unix-like syste…
X8 concepts
Xception ArchitectureNormalization makes samples more similar by centering and scaling their values, which can…XML Attributes and Text ContentXML documents are organized as tree structures with a single root element containing all…XML Basics and StructureXML uses tags and attributes; JSON uses curly braces and key-value pairs to represent the…XML Document Structure and the Root ElementAn XML element is a complete unit consisting of an opening tag, content, and a closing ta…XML Fundamentals and Use CasesJSON has become the industry standard for data exchange because its structure maps direct…XML Structure and SyntaxTriple single quotes (''') and triple double quotes (""") both create multi-line strings…XML Validation and SchemaAn XML element is a complete unit consisting of an opening tag, content, and a closing ta…XML: Structuring Complex DataAn API is an application-to-application contract that publishes rules for accessing servi…
Y2 concepts