Complete self-study package

Modern Deep Learning Foundations

Learn by predicting, running mathematically grounded experiments, explaining evidence, and exporting an inspectable mastery artifact.

Laboratories: 6
Guided time: 4h 20m
Mastery artifacts: 6

0 of 6 laboratories complete Your progress is stored in this browser. Continue this laboratory

Lab 3 of 6 How Does a CNN Work? Intermediate / 50 min / Deliverable: CNN evidence trace

Screen 1 of 3 - Read before you experiment

A CNN transforms spatial evidence one operation at a time

Kernel size defines the local receptive field, while stride and padding determine the output geometry. Convolution creates feature maps, activation keeps useful nonlinear responses, and pooling compresses spatial evidence.

This laboratory connects those geometric rules to real held-out samples and trained CNN tensors. Follow one sample through the pipeline before comparing channels or datasets.

Exact book map

Lesson 8 - How Does a CNN Work?

Printed pages 54-59; PDF pages 56-61.

Open the exact pages

Convolutional Layer: Filters and Sliding Window printed pp. 54-55 / PDF pp. 56-57 filter operation; Equations 8.1-8.2
Padding and Border Effects / Stride printed pp. 55-57 / PDF pp. 57-59 output-geometry paragraphs; Equations 8.3-8.4
Pooling and Receptive Field printed pp. 57-59 / PDF pp. 59-61 pooling and depth discussion; Equation 8.5

Evidence levelReal trained CNNs + held-out MNIST, Fashion-MNIST, and CIFAR-10 samples

What is real, computed, or illustrative

Evidence and provenance

Measured Benchmarks Held-out MNIST, Fashion-MNIST, and CIFAR-10 samples

Images, class labels, and trained-model outputs come from the packaged benchmark artifacts.

Measured Model evidence Learned CNN tensors and probability vectors

Feature maps and predictions are tied to the selected sample and trained network.

Derived Geometry sandbox Exact kernel, stride, padding, pooling, and shape arithmetic

Editable filters demonstrate convolution geometry separately from learned benchmark weights.

Scope Claim boundary Geometry and trained inference are distinct modes

Changing an educational kernel does not retrain the packaged benchmark network.

60-second prerequisite check

Choose the support level from evidence

For a 28 x 28 input, 3 x 3 kernel, stride 1, and no padding, what is the output spatial size? 26 x 26 28 x 28 30 x 30

Use the convolution output-size equation before changing the geometry controls. Padding and stride determine whether borders and resolution are preserved.

Guided task

Complete these three moves

1
Predict
Select a dataset and predict the next tensor shape from kernel size, stride, and padding.
2
Trace
Advance through convolution, activation, and pooling in order. Inspect one feature channel at each stage.
3
Verify
Compare the computed geometry, learned feature maps, probability vector, and final prediction.

Record your prediction before you run the lab

Predict the next tensor height and width from the selected input size, kernel, stride, and padding before revealing the computed shape.

Confidence before experimenting 50% A useful prediction names the expected change and why it should happen.

Start the guided CNN trace

Screen 2 of 3 - Experiment

Run, inspect, and compare

Follow the three guided moves above. Change one variable at a time so every visual change has a clear cause.

Guided mode Predict first, advance one stage at a time, and explain the displayed values before changing another control.

Reproducible experiment record

Compare runs instead of trusting one result

Run the laboratory, then capture the controls, metrics, evidence status, seed, and timestamp.

The accessible visual summary will update when the laboratory renders its first result.

Evidence checkpoint

Explain what happened, then transfer it

0 / 3responses complete

Explain the evidence

Trace one feature channel from convolution through activation and pooling, then connect it to the final probability vector.

Use at least one value or visible change from the experiment.

Transfer the idea

For a new image task, choose a kernel size, stride, and padding policy and explain the spatial information it preserves or discards.

Name the new context and the design choice you would make.

States the input and target spatial scale
Computes the output shape correctly
Justifies kernel, stride, and padding
Explains the information preserved or discarded

Continue to the CNN interpretation

Screen 3 of 3 - Consolidate

By the end of this lesson, you will be able to:

Trace an image through convolution, activation, pooling, and classification layers.
Predict output dimensions from kernel size, stride, and padding settings.
Connect feature-map activations and receptive fields to the final class prediction.

Describes convolution layers, stride, padding, and pooling. Shows how CNNs form feature maps and receptive fields. Explains hierarchical representation of features.

Mastery artifact

Leave with inspectable evidence, not a completion click

Your artifact combines the prerequisite check, prediction, experiment configuration, displayed evidence, explanation, transfer rubric, and confidence change.

Complete the evidence cycle to unlock the artifact.

Import progress JSON

Retention plan

Retrieve the concept after time has passed

Complete the artifact to schedule a 24-hour retrieval prompt and a seven-day transfer revisit.

Learning-study instrument

Help evaluate whether the laboratory teaches the concept

Clarity 3 / 5 Mental effort 3 / 5 Transfer confidence 3 / 5

No name, email address, free text, or IP address is retained in the learning record.