Module 8 · Neural and language models
Module 8: independent implementation practice
Use concepts and guided construction first. Work in the existing module environment. These tasks support the same outcomes as the complete reference; your instructor's assessment requirements remain authoritative.
Start with the clean practice notebook
If a download link does not open, use the offline access route.
- Practice notebook — complete three function bodies and record predictions.
- In a source checkout:
notebooks/practice/module08-practice.ipynb. - The notebook must stay inside the complete source tree and use its installed package/kernel. It does not install dependencies for you.
Run Setup, then define a function and run its check. not_attempted means the placeholder remains; needs_revision supplies feedback; passed covers the visible cases. Repeating a check replaces its earlier state. When all three pass, run the supplied real-development sandbox. Partial work deliberately skips that sandbox rather than pretending it ran.
Implement, check and explain
Task 1: Dense ReLU layer
Implement dense_relu(x, weights, bias) preserving batch-by-hidden shape. Reconcile one activation by hand.
Before checking, write an expected output and explain one likely mistake. After checking, try a changed case and record what the feedback helped you diagnose.
Hint if stuck on task 1
A batch multiplies the input-by-hidden weight matrix; add the hidden bias vector and replace negative activations with zero.
Task 2: Validation checkpoint
Implement select_checkpoint(train_loss, validation_loss). Return a one-based minimum-validation-loss epoch; ties retain the earliest.
Before checking, write an expected output and explain one likely mistake. After checking, try a changed case and record what the feedback helped you diagnose.
Hint if stuck on task 2
The training curve helps interpretation, but its minimum must not choose the checkpoint.
Task 3: Configurable network
Implement build_network(inputs, width) with one ReLU hidden layer and one sigmoid output. Predict the parameter count for two widths.
Before checking, write an expected output and explain one likely mistake. After checking, try a changed case and record what the feedback helped you diagnose.
Hint if stuck on task 3
Count input-hidden weights, hidden biases, hidden-output weights and output bias; batch size is not a learned dimension.
Transfer with less support
In a separate development run, change width from 8 to 4 or 16 and predict the parameter count before training. Keep the data slice, eight-epoch budget and checkpoint rule fixed. Report observed validation loss and compute; a width increase is not guaranteed to help. Use a fresh notebook/kernel or reset all model/optimizer/checkpoint state.
For a hypothetical IT alert queue, describe which false-alarm workload evidence would be needed before preferring a high-recall network. Do not treat the chemistry target as an incident label. Complete the original Module 8 study to compare the actual network with logistic regression on the declared final evaluation.
Evidence for the unchanged outcomes
| Outcome | Evidence to retain |
|---|---|
| M08.1 | Explained activation, probability/loss update and batch/parameter calculations |
| M08.2 | Constructed TensorFlow model and training-only preprocessing in the development scaffold |
| M08.3 | Selected validation epoch and restoration of its actual saved weights |
| M08.4 | Complete-reference neural/logistic confusion comparison with a justified tradeoff |
| M08.5 | Saved/reloaded probabilities, bounded compute record and limitations |
Keep your changed code, one hand-checked numerical claim, expected versus observed output, one diagnosed failure and a short decision memo. Record source/environment/command evidence when using real data. Interpretive reading alone does not replace an implementation outcome.
Get feedback, then revisit
Open practice solutions after your first attempt. Compare the reasoning and a changed case, not only the final number. Return to your original notebook and reproduce the central operation without copying. Use recorded-study feedback after the complete reference run.
If using GenAI, keep a short disclosure of the help used and what you verified. Be able to explain and change the code independently; an equivalent no-assistant route uses the same hints and feedback. Never upload student records or credentials.
Evidence activity: Audit the checkpoint choice
Use the saved training history, validation selection and final predictions. Identify the validation-selected checkpoint, then recalculate one confusion count. Explain why the last epoch or best test score cannot replace the declared checkpoint rule.
Follow the four steps in reproduce a result and justify its use. Submit your artifact path, independent calculation, discrepancy or agreement, and bounded use decision in the evidence worksheet. Use your complete reference run; this activity complements the small function checks.