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 5 of 6 Explainability: Understanding Model Decisions Intermediate / 40 min / Deliverable: Explanation audit

Screen 1 of 3 - Read before you experiment

An explanation highlights evidence; it does not prove causality

Grad-CAM localizes influential regions in a convolutional feature map. Integrated Gradients attributes the prediction relative to a baseline. Occlusion measures how the score changes when image regions are hidden.

Different methods answer different questions and may disagree. First predict where the model should look, then compare the explanation with the object and its surrounding context.

Exact book map

Lesson 15 - Explainability: Understanding Model Decisions

Printed pages 93-99; PDF pages 95-101.

Open the exact pages

Grad-CAM printed pp. 94-95 / PDF pp. 96-97 localization procedure; Equations 15.1-15.3
Permutation Importance and SHAP printed pp. 95-98 / PDF pp. 97-100 feature-removal and additive-attribution paragraphs; Equations 15.4-15.5
Limitations and Trade-offs printed pp. 98-99 / PDF pp. 100-101 closing caution and method-selection paragraphs

Evidence levelAuthentic ResNet-18 predictions + precomputed explanation maps

What is real, computed, or illustrative

Evidence and provenance

Measured Inference ImageNet-pretrained ResNet-18 predictions

Top classes and probabilities were produced for the displayed real images.

Precomputed Explanations Grad-CAM, Integrated Gradients, and Occlusion maps

Attribution assets were generated offline for stable, fast comparison in the browser.

Derived Comparison Method agreement and region-level contrast

The interface compares where methods concentrate evidence for the same target prediction.

Scope Claim boundary Attribution is not causality

A highlighted region may be correlated evidence, background context, or a method artifact.

60-second prerequisite check

Choose the support level from evidence

Two attribution methods highlight different regions. What is the safest conclusion? The explanation is method-dependent and needs a sanity or faithfulness check The model has two causes The brighter map is automatically correct

Attribution disagreement is evidence to investigate. Compare methods, perturb the input, and avoid turning association into a causal claim.

Guided task

Complete these three moves

1
Predict
Choose an object and mark the region you expect to support the prediction.
2
Compare
Inspect Grad-CAM, Integrated Gradients, and Occlusion on the same image.
3
Evaluate
Separate stable object evidence from background correlations and method-specific artifacts.

Record your prediction before you run the lab

Before revealing an explanation, name the object region and background region you expect to support the selected class.

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

Start the explanation comparison

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

Compare the three methods. Identify one stable region, one disagreement, and one conclusion the maps do not justify.

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

Transfer the idea

Design one sanity check or counterfactual image edit that would test whether the explanation is trustworthy.

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

Names a specific explanation method and target class
Defines a counterfactual or perturbation
States the expected score or attribution change
Includes a failure criterion and claim boundary

Continue to explanation reliability

Screen 3 of 3 - Consolidate

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

Interpret positive and negative feature attributions for an individual prediction.
Distinguish a local explanation from claims about global model behaviour.
Assess whether an explanation remains stable under small, label-preserving input changes.

Covers GradCAM, SHAP, and permutation importance. Discusses trust, compliance, and debugging. Notes limitations in resolution, cost, and accuracy.

GradCAM_demo

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.