Beyond linear boundaries
Hidden nonlinearities earn their place by making the final decision simpler in a learned representation.
A nonlinear sensor classifier defeats every classical baseline.
This chapter moves from prove the baseline limit to keep capacity accountable. We will make the idea concrete in a small code example before placing it back inside a production system with a declared latency, cost, and operator-capacity budget.
Compare against the linear baseline on held-out data and add width or depth only when the extra boundary improves the real validation contract.
Verify the behavior illustrated in xor_representation.py, then measure decision utility, slice reliability, and system latency.
An excessive learning rate causes NaNs; dead ReLUs and a BatchNorm train/eval mismatch hide the cause. An unsafe release must degrade to a named baseline or the last known-good version.
Build the idea one layer at a time.
A linear classifier can draw only one flat boundary in its input space. Neural networks become useful when hidden layers reshape that space so a simple final boundary can solve what was previously tangled.
Stacking affine layers without activations does not help: their matrix products collapse into one affine transformation. A nonlinear activation between them creates bends, gates, and reusable regions. The hidden representation is therefore not decoration; it is where an inseparable problem can become separable.
XOR is the smallest honest example. No line separates its two positive corners from its two negative corners in the original plane. Two ReLU units can instead measure which input exceeds the other, and the sum of those hidden coordinates becomes positive exactly for the XOR cases.
- affine layers
- ReLU/GELU
- cross-entropy
- chain rule
Follow the reasoning before touching the code.
- 01
Prove the baseline limit
Plot or enumerate the classes and show why one affine score cannot order every positive above every negative.
- 02
Construct a useful hidden space
Create nonlinear features that respond to different regions, then inspect the hidden coordinates rather than treating them as mysterious neurons.
- 03
Keep capacity accountable
Compare against the linear baseline on held-out data and add width or depth only when the extra boundary improves the real validation contract.
import numpy as np
X = np.array([[0, 0], [0, 1], [1, 0], [1, 1]], dtype=float)
y = np.array([0, 1, 1, 0])
h1 = np.maximum(0, X[:, 0] - X[:, 1])
h2 = np.maximum(0, X[:, 1] - X[:, 0])
hidden = np.column_stack([h1, h2])
prediction = (hidden.sum(axis=1) > 0).astype(int)
print(hidden)
print('correct', prediction.tolist() == y.tolist())Map XOR into a hidden space where one threshold solves it.
What to notice
- The two positive XOR points occupy different hidden axes, but both have hidden sum one; the negative points map to the origin.
- The final rule is linear in hidden space even though the complete input-to-output function is nonlinear.
YOUR TURN Pause and predict before revealing the answer.
Replace each ReLU with its raw affine input. Why can the final sum no longer express XOR?
Need a hint?
Add (x1 - x2) and (x2 - x1) before applying any nonlinearity.
Check your reasoning
The two affine features cancel to zero for every row. Without nonlinear gating, the stacked linear computation collapses and cannot separate XOR.
Classical baselines, Trees & boosting
Connect the chapter to the course’s quantitative spine.
Hidden nonlinearities earn their place by making the final decision simpler in a learned representation. The backpropagated error below is the shared course-level lens that helps quantify the surrounding system. Name each symbol, check its units or shape, and use only values that were available at the production cutoff.
Backpropagated error
| Symbol | Meaning / shape / unit |
|---|---|
δ_l | loss gradient at layer l |
W | weight matrix |
φ′ | activation derivative |
Open derivation and numerical substitution
Start from the production quantity being optimized, substitute the observed values with their declared units, then isolate the model-controlled term. Preserve shape annotations at each step so broadcasting or aggregation cannot silently change the result.
- Write the named inputs: δ_l, W, φ′.
- Substitute one small, hand-checkable batch before vectorizing.
- Calculate an independent reference value and compare within a declared tolerance.
# equation → code contract
inputs = validate_shapes_and_units(batch)
value = compute_c09(inputs)
assert is_finite(value)Change one assumption and make the tradeoff visible.
Use the worked chapter as your prediction, then change one input in the lab and explain why the output moves. The lab runs predefined TypeScript and never executes learner code, so the focus stays on the concept rather than setup.
Optimizer stability lab
Sweep learning rate over a deterministic convex-plus-curvature loss.
Assumption: Twenty gradient steps from w=4 on L(w)=w²/2.
Open nonvisual data table
| Item | Computed state | Interpretation |
|---|---|---|
| step 1 | 3.5200 | 6.19520 |
| step 5 | 2.1109 | 2.22801 |
| step 10 | 1.1140 | 0.62050 |
| step 15 | 0.5879 | 0.17281 |
| step 20 | 0.3103 | 0.04813 |
Trace the complete operating path.
Carry the chapter’s central idea into the system: Hidden nonlinearities earn their place by making the final decision simpler in a learned representation. The path below shows where that reasoning needs an artifact, a test, an owner, or a fallback.
- 01
Turn “Prove the baseline limit” into a written invariant and a failing test.
- 02
Promote xor_representation.py from illustration to a versioned implementation with representative fixtures.
- 03
Use “Keep capacity accountable” as the release checkpoint: Compare against the linear baseline on held-out data and add width or depth only when the extra boundary improves the real validation contract.
- 04
Serve or materialize behind explicit a declared latency, cost, and operator-capacity budget.
- 05
Join telemetry to mature outcomes and retain a rollback path.
Observability
Join service health, input quality, prediction distributions, slice behavior, and mature outcomes by exact version.
Cost
Measure storage, preprocessing, compute, queueing, and human review under a representative arrival pattern.
Failure modes
An excessive learning rate causes NaNs; dead ReLUs and a BatchNorm train/eval mismatch hide the cause. Add a detector, owner, mitigation, and stop condition for this class of failure.
Alternatives
Compare a rule, a simpler statistical baseline, and a different system boundary before adding model complexity.
Explain the contract, not just the vocabulary.
Network from first principles
Specify a NumPy MLP, finite-difference checks, PyTorch parity, calibrated inference, and debugging report.
- Forward/backward derivation
- Gradient check plan
- Optimizer/regularization ablations
- Export and batch-one parity
Read primary material with a purpose.
Return to the opening failure.
Package preprocessing, vocabularies, weights, calibration, output semantics, and batch-one parity as one artifact.
For this lesson, the release evidence is the worked code example, a hand-checked formal result, deterministic simulation output, a ≥80% checkpoint, the production rubric, and a named fallback. The course resolves when the system can produce a small network built from forward pass through backpropagation and optimization.