Standardization
Standardization transforms numerical values by centering and scaling them, commonly to a mean of zero and a standard deviation of one.
Statistics Slam Dunk
Standardization transforms numeric variables to a common scale, commonly by subtracting the mean and dividing by the standard deviation.
- How it is used
- Enables fair comparisons of NBA measures across incompatible scales, seasons, and changing playing conditions.
- How deep it goes
- Defines z-scores, demonstrates grouped R calculations, compares four methods, and interprets standardized NBA results without formal proofs.
- Examples
- Uses payroll versus wins, season-level team statistics, and historical NBA scoring leaders.
- What you can do after
- Apply it via R to create and interpret season-grouped z-scores for comparable numeric analysis.
Machine Learning for Tabular Data: XGBoost, Deep Learning, and AI
A standardized feature is a feature transformed to a normalized scale, commonly with a specified mean and standard deviation.
- How it is used
- Prepares continuous tabular features for classical and deep learning models when their value ranges differ substantially.
- How deep it goes
- Defines mean-and-standard-deviation rescaling and demonstrates pandas, Scikit-learn pipelines, and normalization layers without mathematical derivation.
- Examples
- Uses Auto MPG boxplots, synthetic binary classification, Airbnb NYC feature pipelines, and Kuala Lumpur real-estate columns.
- What you can do after
- Apply it via StandardScaler, pandas calculations, or normalization layers within a preprocessing and model-training workflow.
Ensemble Methods for Machine Learning
Standardization rescales numerical features so that each feature has a specified mean and dispersion, commonly zero mean and unit standard deviation.
- How it is used
- Uses feature scaling to place numerical inputs on comparable ranges before downstream regression or classification.
- How deep it goes
- Explains zero-mean unit-variance scaling, train-only fitting, and library implementation, while deferring broader discrete-feature strategies.
- Examples
- Uses AutoMPG, Bike Sharing, and Australian Credit Approval data to demonstrate numerical scaling alongside regression or categorical encoding.
- What you can do after
- Apply it via sklearn.preprocessing.StandardScaler without fitting preprocessing parameters on the held-out test set.
Read how Ensemble Methods for Machine Learning explains standardization →
Knowledge Graphs and LLMs in Action
Feature standardization transforms numerical features so that they have a specified mean and scale, commonly zero mean and unit variance.
- How it is used
- Prepares heterogeneous node features for downstream classification such as node classification or link prediction.
- How deep it goes
- Demonstrates fitting and reusing StandardScaler, but does not derive the transformation mathematically.
- Examples
- Uses node degree, centrality measures, graph embeddings, and the karate club network as concrete feature examples.
- What you can do after
- Apply it via StandardScaler to align training and prediction features before fitting and evaluating a classifier.
Read how Knowledge Graphs and LLMs in Action explains standardization →
Data Without Labels
Standardization is the transformation of numeric variables to a common scale, typically by centering them and scaling by their variability.
- How it is used
- Prepares differently scaled numeric features for distance-based clustering.
- How deep it goes
- Demonstrates feature selection, z-score and StandardScaler implementations, visualization, and k-means use.
- Examples
- Uses a vehicle dataset, feature scatter plots, and an elbow plot comparing one through nine clusters.
- What you can do after
- Apply standardization with scipy.stats.zscore or sklearn.preprocessing.StandardScaler before fitting k-means.
Build AI Drug Discovery Pipelines
Standardization transforms each feature so that it has a specified mean and scale, commonly mean zero and standard deviation one.
- How it is used
- Prepares differently scaled features for PCA so each feature contributes equally.
- How deep it goes
- Defines mean-zero, unit-standard-deviation scaling and connects it to covariance computation, stopping before implementation detail beyond the PCA workflow.
- Examples
- Uses a 2D coordinate-system illustration and a Python workflow containing StandardScaler and PCA.
- What you can do after
- Apply it via StandardScaler before running PCA on feature data.
Read how Build AI Drug Discovery Pipelines explains standardization →
CUDA for LLMs
Standardization transforms values by subtracting their mean and dividing by their standard deviation.
- How it is used
- Normalizes each token’s feature vector as the standardization step within LayerNorm.
- How deep it goes
- Computes row statistics and discusses fused passes, shared memory, warp reductions, and vectorized loads.
- Examples
- Works through mean, variance, and normalized values for rows containing four features.
- What you can do after
- Build a row-wise LayerNorm kernel that computes statistics and applies normalization and learned scaling.
Also covered in
- Deep Learning with PyTorch, Second Edition
- Financial AI in Practice
- Inside Deep Learning: Math, Algorithms, Models
- Outlier Detection in Python
- Grokking BayesIn early access (MEAP)
- The Well-Grounded Data Analyst
What to know first
- categorical dataCategorical data consists of values representing membership in discrete categories rather than measured numerical magnitudes.
- standard deviationStandard deviation is the nonnegative square root of variance and measures the dispersion of values around their mean in the same units as those values.
- meanThe mean is a measure of central location equal to the arithmetic average of numerical observations, or the probability-weighted average of a random variable’s possible values.
- continuous dataContinuous data consists of numerical values that can vary across a range with arbitrarily fine precision.
Related concepts
- standard deviationuses · Standard deviation is the nonnegative square root of variance and measures the dispersion of values around their mean in the same units as those values.
- StandardScalerimplemented by · StandardScaler is a scikit-learn preprocessing transformer that centers each feature using its fitted mean and scales it using its fitted standard deviation.
- z-scoreproduces · A z-score is a standardized value expressing its signed difference from a reference mean in units of the relevant standard deviation or standard error.
- preprocessingkind of · Preprocessing is the preparation or transformation of data or other inputs into a form suitable for subsequent processing, analysis, or execution.
- feature scalingkind of · Feature scaling is the transformation of numerical features to comparable magnitudes, ranges, or distributions for computation or model fitting.
Questions about standardization
What is standardization?
Standardization transforms numerical values by centering and scaling them, commonly to a mean of zero and a standard deviation of one.
Which Manning books explain standardization?
Statistics Slam Dunk, Machine Learning for Tabular Data: XGBoost, Deep Learning, and AI, Ensemble Methods for Machine Learning, Knowledge Graphs and LLMs in Action, Data Without Labels, Build AI Drug Discovery Pipelines and CUDA for LLMs.
What should I know before learning standardization?
Manning books that teach standardization usually introduce categorical data, standard deviation, mean and continuous data first.
What is standardization a kind of?
Standardization is a kind of preprocessing. Preprocessing is the preparation or transformation of data or other inputs into a form suitable for subsequent processing, analysis, or execution.
What is standardization also called?
Standardization is also called standardize or standardizing.
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.






