liveBook · Manning

Regression

A regression is an unintended deterioration or loss of previously established behavior, performance, or capability following a change.

How 11 Manning books explain it

LLM Customization and Fine-Tuning

A change that makes a model less likely to decline requests outside its competence.

How it is used
Uses regression as a safety and behavioral gate across model adaptation, deployment, and updates.
How deep it goes
Explains failure mechanisms and provides comparative tests, training mitigations, and deployment gates.
Examples
Compares LoRA ranks 16 and 8 on IT support refusals and describes a healthcare model declining fewer out-of-competence questions.
What you can do after
Diagnose regressions by comparing safety responses before and after adaptation, then evaluate mitigations and deployment readiness.

Read how LLM Customization and Fine-Tuning explains regression →

The Art of Unit Testing 3e

A previously working software behavior that becomes broken after a change.

How it is used
Names the risk that a code change breaks previously working functionality.
How deep it goes
Defines the failure and explains why rerunning older tests matters, but defers rapid detection techniques.
Examples
Uses a release-time bug fix that inadvertently breaks an older feature.
What you can do after
Diagnose/debug with it by rerunning tests for previously working features after code changes.

Read how The Art of Unit Testing 3e explains regression →

Machine Learning System Design

A software defect in which a change causes previously working behavior to fail or deteriorate.

How it is used
Uses regression to evaluate whether software, models, or training pipelines remain safe after changes.
How deep it goes
Explains regression through release testing, error tradeoffs, human review, and pipeline checks, without detailing defect-isolation techniques.
Examples
Uses a 100-sample test-set tradeoff and pipeline changes that introduce regression bugs as concrete examples.
What you can do after
Evaluate model and pipeline changes with tests, error tradeoffs, and appropriately assigned human review before release.

Read how Machine Learning System Design explains regression →

Fast by Default

A performance regression is a deterioration in a software system's speed, resource efficiency, or responsiveness after a change.

How it is used
Frames performance changes as user, business, and engineering risks that require measurement and explicit prioritization.
How deep it goes
Explains symptoms, cross-platform consequences, cost modeling, and safeguards, but stops short of prescribing specific profiling implementations.
Examples
Uses messaging lag, mobile checkout slowdown, desktop rendering delays, script accumulation, and regression-cost estimates.
What you can do after
Evaluate performance risk by baselining a critical flow, modeling regression costs, and stating targets or tradeoffs clearly.

Read how Fast by Default explains regression →

Unit Testing Principles, Practices, and Patterns

A regression is a defect in which a feature that previously worked stops working as intended after a change or other event.

How it is used
Frames regression as the quality risk unit tests must guard against during software changes.
How deep it goes
Defines changed behavior after an event, usually code modification, but does not detail regression diagnosis.
Examples
Uses feature changes, refactoring, releases, and tests as contexts where regressions appear or are prevented.
What you can do after
Evaluate whether tests provide confidence that changes will not break existing functionality.

Read how Unit Testing Principles, Practices, and Patterns explains regression →

Vibe Engineering

A regression is an unintended loss or degradation of previously correct software behavior.

How it is used
Uses regression as a review and benchmarking concern when changes may degrade performance, compatibility, security, or other established behavior.
How deep it goes
Demonstrates detection through benchmarks and review gates, but stops short of a general taxonomy or formal regression-testing method.
Examples
Uses AccountFinder concurrency variants, custom timing results, JMH benchmarks, and Git-based AI review hooks.
What you can do after
Diagnose regressions by combining deterministic gates, scoped review, result-consistency checks, and representative performance benchmarks.

Read how Vibe Engineering explains regression →

AI Model Evaluation

A regression is an unintended deterioration in a model or system behavior after a change.

How it is used
Protects model launches and evaluations from shipping quality or infrastructure changes that quietly worsen user or system outcomes.
How deep it goes
Traces suspected failures through data, judges, prompts, rubrics, stability checks, models, and infrastructure guardrails.
Examples
Uses falling win rates, factuality scores, failed experiment gates, changed canary scores, latency, throughput, cost, and downstream errors.
What you can do after
Diagnose whether an apparent regression comes from data, evaluation machinery, model behavior, or infrastructure before escalating it.

Read how AI Model Evaluation explains regression →

Test Driven: Practical TDD and Acceptance TDD for Java Developers

In software, a change from previously working functionality to broken functionality.

How it is used
Uses regression testing to check that software changes do not break functionality that previously worked.
How deep it goes
Defines regression and outlines repeated automated checks, but does not detail their implementation.
Examples
Uses a test-harness mold metaphor to show how a breaking change signals a regression.
What you can do after
Operate an automated test harness repeatedly to detect failures as software changes.

Read how Test Driven: Practical TDD and Acceptance TDD for Java Developers explains regression →

Retrieval Augmented Generation, The Foundational Ideas

A regression is a deterioration in system behavior caused by a change that previously improved or preserved performance.

How it is used
Uses regression tracking to prevent unmeasured deterioration while iterating on production RAG pipelines.
How deep it goes
Explains interacting pipeline causes, trend detection, version diffs, targeted fixes, deployment, and monitoring.
Examples
Uses a mortgage compliance assistant, a cross-encoder reranker, successive evaluation reports, and a RAG pipeline version manifest.
What you can do after
Diagnose regressions by comparing evaluation trends, diffing pipeline versions, and applying targeted configuration or model changes.

Read how Retrieval Augmented Generation, The Foundational Ideas explains regression →

Designing Deep Learning Systems: A guide for software engineers

A deterioration in a model's measured performance relative to an earlier version or evaluation.

How it is used
Uses regression to motivate reproducible troubleshooting and post-deployment monitoring of production models.
How deep it goes
Explains regression as a detection and troubleshooting concern, but does not specify a detection algorithm or diagnostic procedure.
Examples
Uses fraud detection, loan approval, chatbot routing, and competing production model versions as operational examples.
What you can do after
Diagnose/debug regression by reproducing models, comparing datasets and algorithm code, and monitoring deployed versions.

Read how Designing Deep Learning Systems: A guide for software engineers explains regression →

Secrets of the JavaScript Ninja, Second Edition

A failure of previously working software behavior following a change.

How it is used
Frames browser regressions as a major risk to reusable, maintainable JavaScript across browsers.
How deep it goes
Defines regressions, distinguishes predictable API changes from unpredictable bugs, and recommends feature detection, testing, and release monitoring.
Examples
Uses the Internet Explorer transition from attachEvent to addEventListener to demonstrate feature detection.
What you can do after
Diagnose regressions by testing browser releases and addressing failures; anticipate known API changes with feature detection.

Read how Secrets of the JavaScript Ninja, Second Edition explains regression →

Also covered in

What to know first

Related concepts

Questions about regression

What is regression?

A regression is an unintended deterioration or loss of previously established behavior, performance, or capability following a change.

Which Manning books explain regression?

LLM Customization and Fine-Tuning, The Art of Unit Testing 3e, Machine Learning System Design, Fast by Default, Unit Testing Principles, Practices, and Patterns, Vibe Engineering, AI Model Evaluation, Test Driven: Practical TDD and Acceptance TDD for Java Developers and 3 more.

What should I know before learning regression?

Manning books that teach regression usually introduce software testing, language model, acceptance criteria and foundation model first.

What is the difference between regression and data drift?

A regression is an unintended deterioration or loss of previously established behavior, performance, or capability following a change. Data drift is a change in the statistical distribution or characteristics of data relative to a reference, such as data used to train a model.

What kinds of regression are there?

Kinds of regression include linear regression.

These guides are generated from the text of Manning books by Manning's knowledge graph and link to the sections they draw on. The full books are available on liveBook and at manning.com.