Optimization

Batch Normalization

Normalize activations using batch statistics to stabilize deep training (and understand the train vs inference mismatch).

status: reviewimportance: importantdifficulty 3/5math: undergraduateread: 14mdemo planned

Back to Optimization Next: Layer Normalization & RMSNorm

Concept Structure

Batch Normalization

01Intuition

Start with the picture, metaphor, or geometric mechanism.

02Math

Make the objects explicit and connect them with notation.

03Code

Mirror the equations with runnable implementation details.

04Interactive Demo

Manipulate the mechanism and watch the idea respond.

1prerequisites

1next concepts

2related links

Learning map

Batch Normalization

BeforeBackpropagationNow3/4 sections readyTryUse the demo notes to predict the mechanism before moving on.NextLayer Normalization & RMSNorm

Object flow

3/4 sections readyAsk about this Research room

ConceptBatch NormalizationOptimization EquationBatch Normalization equation 1Exact equation object CodeBatch Normalization code witness 1Exact code witness ClaimBatch Normalization central claimMechanism claim EquationBatch Normalization math objectsEquation and notation CodeBatch Normalization code witnessRunnable code object

ConceptBatch NormalizationOptimization

Local snapshot ready

concept:optimization/batch-normalization

Codewitness nearby Predictbefore reveal Roomobject handoff

Conceptual Bridge

What should feel connected as you move through this page.

Carry inBackpropagation

Bring the mental model from Backpropagation; this page will reuse it instead of restarting from zero.

Work hereBatch Normalization

Normalize activations using batch statistics to stabilize deep training (and understand the train vs inference mismatch).

Carry outLayer Normalization & RMSNorm

The next edge should feel earned: use the demo prediction here before following Layer Normalization & RMSNorm.

Test the linkUse the demo notes to predict the mechanism before moving on.Then continue to Layer Normalization & RMSNorm

01IntuitionStart with the picture, metaphor, or geometric mechanism.02MathMake the objects explicit and connect them with notation.03CodeMirror the equations with runnable implementation details.04Interactive DemoManipulate the mechanism and watch the idea respond.

Intuition

Build the mental picture first so the rest of the page has something to attach to.

Section prompt

BatchNorm was one of the key tricks that made very deep convolutional networks train reliably.

The basic idea is simple: for each feature channel, normalize activations using the mean and variance computed over the current mini-batch. This keeps activations in a stable numeric range and often makes optimization easier.

But BatchNorm has an important downside: it introduces a train vs inference mismatch. During training you use batch statistics; at inference you use running averages. This is one reason transformers prefer LayerNorm/RMSNorm.

Math

Translate the story into symbols, assumptions, and a derivation you can inspect.

Section prompt

Equation 1\hat x_{b,j} = \frac{x_{b,j} - \mu_j}{\sqrt{\sigma_j^2 + \epsilon}},Equation 2\mu_j = \frac{1}{m}\sum_{b=1}^m x_{b,j},\qquad \sigma_j^2 = \frac{1}{m}\sum_{b=1}^m (x_{b,j}...

For a batch of activations $x\in\mathbb R^{m\times d}$ (batch size $m$ , features $d$ ), BatchNorm normalizes each feature dimension:

\hat x_{b,j} = \frac{x_{b,j} - \mu_j}{\sqrt{\sigma_j^2 + \epsilon}},

where

\mu_j = \frac{1}{m}\sum_{b=1}^m x_{b,j},\qquad \sigma_j^2 = \frac{1}{m}\sum_{b=1}^m (x_{b,j} - \mu_j)^2.

Then apply learned scale and shift:

y_{b,j} = \gamma_j\,\hat x_{b,j} + \beta_j.

The learned $\gamma_j$ and $\beta_j$ are important: normalization gives the optimizer a stable coordinate system, but the model can still recover whatever feature scale and offset are useful. During backpropagation, each example's normalized value depends on the other examples in the same mini-batch through $\mu_j$ and $\sigma_j^2$ , so the layer is not example-independent while training.

At inference, you typically use running averages of $(\mu_j,\sigma_j^2)$ accumulated during training instead of per-batch values.

Code

Keep the implementation aligned with the notation so the algorithm is legible.

Section prompt

Code witness 1import numpy as np rs = np.random.RandomState(0) x = rs.randn(6, 4) * 3.0 + 5.0 # batch activ...python

import numpy as np

rs = np.random.RandomState(0)
x = rs.randn(6, 4) * 3.0 + 5.0  # batch activations with nonzero mean/scale
eps = 1e-5

mu = x.mean(axis=0, keepdims=True)
var = ((x - mu) ** 2).mean(axis=0, keepdims=True)
xhat = (x - mu) / np.sqrt(var + eps)

print("per-feature mean before:", np.round(x.mean(axis=0), 3))
print("per-feature mean after :", np.round(xhat.mean(axis=0), 3))
print("per-feature var after  :", np.round(xhat.var(axis=0), 3))

Interactive Demo

Use direct manipulation to connect the explanation to a moving system.

Section prompt

No interactive demo yet for this concept. A good next step is a small visualization that shows:

how batch size affects the noise in batch statistics,
why small batches can destabilize BatchNorm,
why LayerNorm avoids the batch coupling.

No live visualization is registered for this concept yet.

The page still supports explanatory demo notes above; when a viz.tsx exists, it will render here without changing the route.

After The First Pass

Turn the concept into an inspected object.

Once the invariant is visible in the intuition, math, code, and demo, use these panels to inspect the mechanism visually, check source support, practice the idea, and attach a grounded research question.

Mechanism Storyboard

See the idea move before the page explains it

Normalize activations using batch statistics to stabilize deep training (and understand the train vs inference mismatch).

Demo notes open01 / Intuition

Prediction lens

Start with the picture, metaphor, or geometric mechanism.

Commit first

Before reading further, choose the kind of change Batch Normalization should make visible.

Visual Inquiry

Make the image answer a mathematical question

Normalize activations using batch statistics to stabilize deep training (and understand the train vs inference mismatch).

3/4 stages readyDemo notes connected

Visual cueWhich visible object should carry the first intuition?

Inspection depth2/4

Prediction

Which visible object should carry the first intuition?

Commit first

Pick the cue that should make Batch Normalization easier to reason about before the page gives the answer.

Claim Review

Normalize activations using batch statistics to stabilize deep training (and understand the train vs inference mismatch).

StatusSubstantive claim review pending

Source IDs and witness objects are attached for review; they are not proof by themselves.

SourcesNo references

Add source metadata before claiming support.

Witnesses3 local objects

Use equation, code, and demo objects to check whether the source support is operational.

Source support candidates

No structured source note is attached yet.

Mechanism witnesses

Equation 1

\hat x_{b,j} = \frac{x_{b,j} - \mu_j}{\sqrt{\sigma_j^2 + \epsilon}},

Equation 2

\mu_j = \frac{1}{m}\sum_{b=1}^m x_{b,j},\qquad \sigma_j^2 = \frac{1}{m}\sum_{b=1}^m (x_{b,j} - \mu_j)^2.

Code witness 1import numpy as np rs = np.random.RandomState(0) x = rs.randn(6, 4) * 3.0 + 5.0 # batch activ...

Practice Loop

Try the idea before it explains itself

Normalize activations using batch statistics to stabilize deep training (and understand the train vs inference mismatch).

Readiness0/3 checks ready

Predict

Before touching the demo, predict one visible change that should happen in Batch Normalization.

Hint 1

Reveal when your model needs a nudge.

Hint 2

Reveal when your model needs a nudge.

Hint 3

Reveal when your model needs a nudge.

Your answer canvas

Local checks

Claim

A concrete answer is on the canvas.

Mechanism

The answer names why the claim should hold.

Bridge

It touches the page context or a neighboring idea.

Misconception check

Object research drawerClose

ConceptBatch NormalizationOptimization

Code witness comparisonBatch Normalization code witness 1rs = np.random.RandomState(0)Prediction before revealBatch Normalization predictionUse the demo notes to predict the mechanism before moving on.

Grounded room questionWhat is the smallest example that makes Batch Normalization click without losing the math?Local snapshot ready

Research Room

Attach the question to an exact object

Pick the concept, equation, source, code witness, claim, misconception, or demo state before asking for help. The handoff stays grounded to that object.

Next local actionNo local draft saved yet

Open the draft below to save one note and next action in this browser.

conceptOptimization

Batch Normalization

Anchored question

What is the smallest example that makes Batch Normalization click without losing the math?

Local action draftNo local draft saved yetExpand only when ready to capture one local next action

Local action draft

This draft stays locally in this browser for concept:optimization/batch-normalization.

Draft noteNext action

No local draft saved.

Evidence to inspect

Definition, prerequisite, and contrast concept links
The equation or code witness that makes the concept operational
One demo state that shows the invariant instead of a slogan

What would resolve this

The learner can state the mechanism in their own words
The learner can name the prerequisite that would repair confusion
The learner can predict how the mechanism changes under one perturbation

Grounded AI handoff

I am working in Continuous Function's research reading room. Object: concept - Batch Normalization Object key: concept:optimization/batch-normalization Context: Optimization Anchor id: concept/concept-notebook/optimization/batch-normalization Open question: What is the smallest example that makes Batch Normalization click without losing the math? Evidence to inspect: - Definition, prerequisite, and contrast concept links - The equation or code witness that makes the concept operational - One demo state that shows the invariant instead of a slogan What would resolve this: - The learner can state the mechanism in their own words - The learner can name the prerequisite that would repair confusion - The learner can predict how the mechanism changes under one perturbation Answer as a careful research tutor: stay source-grounded, separate verified evidence from assumptions, name the relevant math objects, and end with one next action.

Open source object

concept/concept-notebook/optimization/batch-normalization
concept:optimization/batch-normalization

Learning Map

Before / Now / Try / Next

BeforeBackpropagation

NowIntuition → Math → Code → Demo

TryUse the demo notes to predict the mechanism before moving on.

NextLayer Normalization & RMSNorm

Intuitionready
Mathready
Codeready
Interactive Demoplanned

Object Companion

Ask beside the selected object

Normalize activations using batch statistics to stabilize deep training (and understand the train vs inference mismatch).

Your question

GoalComfortStyleStuck on

Context prompt

You are my AI learning companion for Continuous Function. Current context: Optimization concept. Learning surface: Batch Normalization. What this page says: Normalize activations using batch statistics to stabilize deep training (and understand the train vs inference mismatch). Current section: Intuition, math, code, and interactive demo. Suggested next step: Use the demo notes to predict the mechanism before moving on.. Learner goal: Understand the idea. Learner comfort level: New to this. Preferred explanation style: Visual first. Task: Explain the central idea in plain language, then restate it with the exact math objects from the page. Answer in a way that helps me learn: ask one clarifying question only if needed, use intuition before notation, and end with one thing I should try on the page.

Domain

Optimization

optimizationbatch-normalizationtraining-stability

Prerequisites

Backpropagation

Leads To

Layer Normalization & RMSNorm

Weight Initialization: Xavier, He & muP Learning Rate Schedules: Warmup, Decay & Cycling

Within this domain