# Direct Assessment Evidence Template: AINS6003 Deep Learning & Neural Networks

## Purpose

This template defines the evidence Aurnova should retain after each offering of **AINS6003 Deep Learning & Neural Networks**. It supports accreditation review, annual assessment, faculty calibration, and continuous improvement. It should be completed by the instructor or program assessment lead after grades are finalized.

## Offering Information

| Field | Entry |
|---|---|
| Term / Cohort | TBD |
| Instructor of Record | TBD |
| Delivery Mode | Online 8-week course |
| Enrollment | TBD |
| Completion Rate | TBD |
| Date Reviewed by Program | TBD |

## Signature Assessment Evidence

| Module | Topic | Direct Evidence to Retain | Outcome Alignment | Sample / Report Location |
|---|---|---|---|---|
| 1 | From neurons to multilayer networks | Module 1 Assignment: Baseline MLP design brief; notebook lab; rubric score; reflection | CO1, CO2, CO6 | High/mid/low artifact samples; rubric summary; instructor notes |
| 2 | Backpropagation and automatic differentiation | Module 2 Assignment: Gradient trace and autograd check; notebook lab; rubric score; reflection | CO2, CO3, CO6 | High/mid/low artifact samples; rubric summary; instructor notes |
| 3 | Optimization, loss, and regularization | Module 3 Assignment: Optimizer and regularization comparison; notebook lab; rubric score; reflection | CO2, CO3, CO6 | High/mid/low artifact samples; rubric summary; instructor notes |
| 4 | Convolutional neural networks for vision | Module 4 Assignment: CNN architecture memo; notebook lab; rubric score; reflection | CO3, CO4, CO6 | High/mid/low artifact samples; rubric summary; instructor notes |
| 5 | Sequence models: RNNs and LSTMs | Module 5 Assignment: Sequence modeling design note; notebook lab; rubric score; reflection | CO3, CO5, CO6 | High/mid/low artifact samples; rubric summary; instructor notes |
| 6 | Attention and transformers | Module 6 Assignment: Attention mechanism analysis; notebook lab; rubric score; reflection | CO4, CO5, CO6 | High/mid/low artifact samples; rubric summary; instructor notes |
| 7 | Generative models and applications | Module 7 Assignment: Discriminative vs generative comparison; notebook lab; rubric score; reflection | CO4, CO5, CO6 | High/mid/low artifact samples; rubric summary; instructor notes |
| 8 | GPU workflows, scale, and deployment | Module 8 Assignment: Reproducible training workflow checklist; notebook lab; rubric score; reflection | CO1, CO4, CO6 | High/mid/low artifact samples; rubric summary; instructor notes |

## Rubric Summary Table

| Outcome | Number of Students Assessed | % Excellent | % Proficient | % Developing | % Not Yet Demonstrated | Action Needed? |
|---|---:|---:|---:|---:|---:|---|
| CO1 | TBD | TBD | TBD | TBD | TBD | TBD |
| CO2 | TBD | TBD | TBD | TBD | TBD | TBD |
| CO3 | TBD | TBD | TBD | TBD | TBD | TBD |
| CO4 | TBD | TBD | TBD | TBD | TBD | TBD |
| CO5 | TBD | TBD | TBD | TBD | TBD | TBD |
| CO6 | TBD | TBD | TBD | TBD | TBD | TBD |

## Artifact Sampling Protocol

Retain anonymized or access-controlled samples representing high, satisfactory, and needs-revision performance for each major assessment category when available. Each sample should include the student submission, completed rubric, instructor feedback, and revision history if applicable. Do not include personally identifiable information in public repositories or accreditor packets unless institutional policy authorizes it.

## Faculty Calibration Record

| Calibration Item | Evidence |
|---|---|
| Sample artifact reviewed before grading | TBD |
| Faculty or reviewer participants | TBD |
| Rubric interpretation issue identified | TBD |
| Resolution / scoring norm | TBD |
| Follow-up needed in next offering | TBD |

## Continuous Improvement Log

| Evidence Reviewed | Finding | Improvement Action | Owner | Due Date | Status |
|---|---|---|---|---|---|
| Grade distribution | TBD | TBD | TBD | TBD | TBD |
| Rubric outcome performance | TBD | TBD | TBD | TBD | TBD |
| Student feedback | TBD | TBD | TBD | TBD | TBD |
| Instructor reflection | TBD | TBD | TBD | TBD | TBD |
| Accessibility / technology friction | TBD | TBD | TBD | TBD | TBD |

## Program Assessment Roll-Up

After this course-level template is completed, the instructor or program assessment lead should roll the evidence into the AINS program assessment cycle documented in `CastaliaInstitute/MSAI/accreditation/AINS-program-assessment-cycle.md`. The roll-up should identify which PLO benchmarks were met, which course outcomes need attention, what evidence was sampled, and what improvement action will be checked in the next offering. This is the course-level source for annual assessment, curriculum-map review, benchmark analysis, and closing-the-loop documentation.

## Accreditation Notes

This template is intentionally course-level. Institution-level records such as faculty credentials, disability accommodation records, official gradebooks, identity verification, and LMS activity reports should remain in Aurnova-controlled systems and be referenced from the central accreditation portfolio rather than stored in public course repositories.