JigyasaWorkshop 01 / Learning workspace
One place for the teaching day. Open the slides, run the next practical, record your evidence and keep your work.
Third-year engineering · Basic Python · 09:00–17:00
The host confirms the event date, venue and lab arrangements.
Before the day
Extract the learner ZIP. Keep data/ beside notebooks/. Install the requirements before class, then open JupyterLab.
python -m pip install -r requirements.txt python -m jupyterlab .Requirements · Setup instructions
Open Colab, choose File → Upload notebook and select a downloaded notebook. Upload the CSV through Files. If no local CSV exists, the notebook downloads the public synthetic dataset.
A Google sign-in and working internet may be required. A rehearsal on the host lab and network remains necessary.
The practical pathway
Each notebook starts from the same CSV and can run independently. Guided cells work as written. Complete the response prompts and extensions yourself.
10:00–10:45 / Practical 01
Inspect the CSV and identify which columns can exist at prediction time.
Draw an input/target table. Explain why the ID is excluded. Count missing values on training and describe how validation uses training medians.
Explain one result to your partner, swap keyboard roles and save the notebook.
11:00–12:00 / Practical 02
Predict numerical energy and compare against a training-mean baseline.
Calculate MAE by hand for actual [100,120,140] and predicted [110,115,135]. Try occupancy-only regression using the same validation rows. Record whether all four inputs improve MAE.
Explain one result to your partner, swap keyboard roles and save the notebook.
12:00–12:45 / Practical 03
Classify high demand with a probability model.
Explain the difference between 125 kWh label definition and 0.5 probability threshold. Record a false negative and false positive scenario.
Explain one result to your partner, swap keyboard roles and save the notebook.
13:30–14:15 / Practical 04
Choose a decision threshold using validation evidence.
For TP=18, FP=6, FN=2, TN=74 compute accuracy, precision and recall. Change missed-event cost from 5 to 2 using validation only.
Explain one result to your partner, swap keyboard roles and save the notebook.
14:15–15:00 / Practical 05
Compare tree complexity and the training/validation gap.
Choose a tree depth from the validation table and justify it. Explain why maximum training F1 is insufficient. Sketch the first two decision rules.
Explain one result to your partner, swap keyboard roles and save the notebook.
15:15–16:15 / Practical 06
Audit a suspicious result, freeze decisions and report final evidence.
Write a model card with decision, inputs, exclusions, split, validation choice, test metrics, failure and limitation. Present why near-zero MAE was invalid.
Explain one result to your partner, swap keyboard roles and save the notebook.
NameError: restart the kernel and run from the top. Missing package: use the correct environment and requirements. CSV not found: extract the ZIP or upload the CSV to Colab. NaN error: run the full Pipeline cell. Different score: check the seed, inputs and data split before changing the model.
No network? Use the local CSV and preinstalled Jupyter. If a device cannot run, work with a partner and record that limitation honestly.
Teaching screen
A focused 26-slide sequence adapted from Akash Chatake’s college ML teaching topics. The editable PowerPoint contains public facilitation notes.
The listening shelf
Selected existing teaching episodes for preparation and revision. Use headphones outside the timed practical blocks. The topic notes below are reading guides, not verbatim transcripts.
10 min 34 sec / College ML teaching collection
Download episodeListening prompt: connect this topic to the matching practical and write one question for discussion.
12 min 47 sec / College ML teaching collection
Download episodeListening prompt: connect this topic to the matching practical and write one question for discussion.
6 min 30 sec / College ML teaching collection
Download episodeListening prompt: connect this topic to the matching practical and write one question for discussion.
15 min 17 sec / College ML teaching collection
Download episodeListening prompt: connect this topic to the matching practical and write one question for discussion.
18 min 22 sec / College ML teaching collection
Download episodeListening prompt: connect this topic to the matching practical and write one question for discussion.
The teaching clock
| Time | Focus | Teaching cue |
|---|---|---|
| 09:00–09:05 | Readiness and engineering judgement | Welcome and define the campus-energy decision |
| 09:05–09:10 | Readiness and engineering judgement | Ask what Python each learner has used |
| 09:10–09:15 | Readiness and engineering judgement | Pair students and appoint driver/navigator |
| 09:15–09:20 | Readiness and engineering judgement | Open workspace and downloaded notebook pack |
| 09:20–09:25 | Readiness and engineering judgement | Run import check; route failed machines to pairs |
| 09:25–09:30 | Readiness and engineering judgement | Ask two students to distinguish prediction from action |
| 09:30–09:35 | ML workflow | Contrast rules, learned model and generative AI |
| 09:35–09:40 | ML workflow | Ask whether the target is a number or a category |
| 09:40–09:45 | ML workflow | Show the four input columns and prediction time |
| 09:45–09:50 | ML workflow | Draw training/validation/test roles |
| 09:50–09:55 | ML workflow | Explain the independent synthetic-scenario assumption |
| 09:55–10:00 | ML workflow | Get each pair to name one forbidden future input |
| 10:00–10:05 | Lab 01 data | Open notebook01 and read the data card |
| 10:05–10:10 | Lab 01 data | Run setup; confirm 360/120/120 |
| 10:10–10:15 | Lab 01 data | Inspect the training head and units |
| 10:15–10:20 | Lab 01 data | Count missing values on training |
| 10:20–10:25 | Lab 01 data | Ask why zero and unknown differ |
| 10:25–10:30 | Lab 01 data | Run training-median imputation |
| 10:30–10:35 | Lab 01 data | Audit excluded target/ID columns |
| 10:35–10:40 | Lab 01 data | Swap drivers; write the input table |
| 10:40–10:45 | Lab 01 data | Checkpoint: every pair explains one preprocessing choice |
| 10:45–10:50 | Morning break | Save notebooks and step away |
| 10:50–10:55 | Morning break | Check two stuck machines privately |
| 10:55–11:00 | Morning break | Announce five-minute return reminder |
| 11:00–11:05 | Lab 02 regression | Explain numerical prediction and baseline |
| 11:05–11:10 | Lab 02 regression | Open notebook02 in a fresh kernel |
| 11:10–11:15 | Lab 02 regression | Run setup and training-mean baseline |
| 11:15–11:20 | Lab 02 regression | Predict whether four features will help |
| 11:20–11:25 | Lab 02 regression | Fit linear regression |
| 11:25–11:30 | Lab 02 regression | Read validation MAE in kWh |
| 11:30–11:35 | Lab 02 regression | Inspect observed/predicted plot |
| 11:35–11:40 | Lab 02 regression | Compute three-error MAE by hand |
| 11:40–11:45 | Lab 02 regression | Switch to occupancy-only input as extension |
| 11:45–11:50 | Lab 02 regression | Compare the same validation rows |
| 11:50–11:55 | Lab 02 regression | Swap drivers and write conclusion |
| 11:55–12:00 | Lab 02 regression | Checkpoint: model and baseline with units |
| 12:00–12:05 | Lab 03 classification | Explain observed high-demand label |
| 12:05–12:10 | Lab 03 classification | Contrast label threshold and probability threshold |
| 12:10–12:15 | Lab 03 classification | Run logistic pipeline |
| 12:15–12:20 | Lab 03 classification | Inspect predicted probabilities |
| 12:20–12:25 | Lab 03 classification | Convert probabilities at 0.5 |
| 12:25–12:30 | Lab 03 classification | Compare majority baseline |
| 12:30–12:35 | Lab 03 classification | Find a false positive |
| 12:35–12:40 | Lab 03 classification | Find a false negative; swap drivers |
| 12:40–12:45 | Lab 03 classification | Checkpoint: explain an actual classification error |
| 12:45–12:50 | Lunch | Save work and announce return time |
| 12:50–12:55 | Lunch | Save work and announce return time |
| 12:55–13:00 | Lunch | Save work and announce return time |
| 13:00–13:05 | Lunch | Save work and announce return time |
| 13:05–13:10 | Lunch | Save work and announce return time |
| 13:10–13:15 | Lunch | Save work and announce return time |
| 13:15–13:20 | Lunch | Save work and announce return time |
| 13:20–13:25 | Lunch | Save work and announce return time |
| 13:25–13:30 | Lunch | Save work and announce return time |
| 13:30–13:35 | Lab 04 evaluation | Reconstruct the four confusion counts |
| 13:35–13:40 | Lab 04 evaluation | Calculate accuracy by hand |
| 13:40–13:45 | Lab 04 evaluation | Calculate precision and recall |
| 13:45–13:50 | Lab 04 evaluation | Run threshold sweep on validation |
| 13:50–13:55 | Lab 04 evaluation | Introduce classroom cost 5 FN + 1 FP |
| 13:55–14:00 | Lab 04 evaluation | Choose smallest cost with documented tie-break |
| 14:00–14:05 | Lab 04 evaluation | Try cost 2 FN + 1 FP if ready |
| 14:05–14:10 | Lab 04 evaluation | Write why test data cannot choose threshold |
| 14:10–14:15 | Lab 04 evaluation | Checkpoint: explain the chosen threshold and assumption |
| 14:15–14:20 | Lab 05 trees | Explain tree question and leaf prediction |
| 14:20–14:25 | Lab 05 trees | Run depth1 and depth2 |
| 14:25–14:30 | Lab 05 trees | Run depth3 and depth5 |
| 14:30–14:35 | Lab 05 trees | Compare unrestricted depth |
| 14:35–14:40 | Lab 05 trees | Read training and validation F1 together |
| 14:40–14:45 | Lab 05 trees | Discuss overfitting and sample uncertainty |
| 14:45–14:50 | Lab 05 trees | Sketch first two rules or use export_text |
| 14:50–14:55 | Lab 05 trees | Choose a defensible depth |
| 14:55–15:00 | Lab 05 trees | Checkpoint: evidence instead of maximum training score |
| 15:00–15:05 | Afternoon break | Save all results |
| 15:05–15:10 | Afternoon break | Check challenge readiness |
| 15:10–15:15 | Afternoon break | Call learners back for investigation |
| 15:15–15:20 | Lab 06 leakage and final test | Introduce suspicious meter feature |
| 15:20–15:25 | Lab 06 leakage and final test | Ask pairs to predict score before running |
| 15:25–15:30 | Lab 06 leakage and final test | Run deliberately invalid regression |
| 15:30–15:35 | Lab 06 leakage and final test | Audit feature availability at decision time |
| 15:35–15:40 | Lab 06 leakage and final test | Remove meter and name correct inputs |
| 15:40–15:45 | Lab 06 leakage and final test | Rebuild safe training-only preprocessing |
| 15:45–15:50 | Lab 06 leakage and final test | Select logistic threshold on validation |
| 15:50–15:55 | Lab 06 leakage and final test | Write frozen selection.json |
| 15:55–16:00 | Lab 06 leakage and final test | Open held-out test once |
| 16:00–16:05 | Lab 06 leakage and final test | Record confusion counts and regression MAE |
| 16:05–16:10 | Lab 06 leakage and final test | Complete model card and limitations |
| 16:10–16:15 | Lab 06 leakage and final test | Checkpoint: preserve disappointing results honestly |
| 16:15–16:20 | Evidence review | Choose three pairs with different errors |
| 16:20–16:25 | Evidence review | First pair: decision and baseline |
| 16:25–16:30 | Evidence review | Second pair: threshold and confusion matrix |
| 16:30–16:35 | Evidence review | Third pair: leakage diagnosis |
| 16:35–16:40 | Evidence review | Each student writes their individual contribution |
| 16:40–16:45 | Exit and next steps | Hand out five exit questions |
| 16:45–16:50 | Exit and next steps | Quiet individual response |
| 16:50–16:55 | Exit and next steps | Explain seven-day extension and feedback |
| 16:55–17:00 | Exit and next steps | Verify attendance and participation process; close |
A local activity aid. It does not change the workshop schedule.
Print desk
Nine pages covering the workflow, practical explanations, data assumptions and seven-day extension.
Open notes PDFFour pages: data/regression, classification, model card, and individual exit response.
Open worksheet PDFSuggested print count: one four-page worksheet per student. Notes can remain digital; print one per pair only if needed. Keep one faculty runbook and attendance sheet with the organiser.
Faculty access
The private faculty pack contains five original college unit decks, reference practicals, expected outputs, the full runbook and participation templates. The complete ZIP, five directly viewable college decks, runbook, expected results and blank participation documents are available in the restricted faculty Drive. Sign in with the authorised faculty account.
Google Drive permission controls access. Sign in with the account that owns the faculty folder. A public link to this folder does not make its contents public.
Your five complete unit presentations, preserved with their original design. Open PowerPoint for the editable deck or PDF for quick viewing. These links require the authorised Google Drive account. The separate 26-slide workshop guide remains under Lecture.
Original college deck · Unit 1 · 30 slides
Opening: slides 3–10 and 29. Explain learning, prediction types and the workflow.
Original college deck · Unit 2 · 30 slides
Before the data practical: slides 7–10, 15–25. Cover missing values, scaling, encoding and splits. Fit transformations on training data only.
Original college deck · Unit 3 · 30 slides
Feature discussion: slides 2–5. Keep PCA and LDA as optional follow-up.
Original college deck · Unit 4 · 30 slides
Main teaching block: slides 5–12, 17–26. Trees, KNN, regression, classification and evaluation.
Original college deck · Unit 5 · 35 slides
Closing extension: slides 2–9 and 26–30. Clustering, responsible use and the path towards deployment.
The organiser verifies attendance, confirms the event date and approves names before issuing certificates. Templates remain unissued and private. This page does not generate certificates or collect student records.
Reproduce the experiment, investigate a failure and write a model card over the next seven days.