Accreditation Evidence Packet: AINS6003 Deep Learning & Neural Networks

Accreditation Status Statement

This packet documents the course-level evidence needed for accreditation review, internal quality assurance, and continuous improvement. The course is designed as a 3-credit, 8-week graduate course in the Aurnova Master of Science in Artificial Intelligence. It supports students who are interested in AI but may not have undergraduate preparation in computer science.

Program Learning Outcomes Used for Alignment

PLO	Name	Description
PLO1	AI Foundations and Methods	Apply core AI, machine learning, and data concepts to professional problems.
PLO2	Technical and Data Practice	Use reproducible computational workflows, data preparation, and evaluation practices appropriate to the task.
PLO3	Responsible AI and Risk	Assess ethical, legal, privacy, security, fairness, and operational risks in AI-enabled systems.
PLO4	Deployment and Operations	Plan AI projects, deployment pathways, monitoring, governance, and lifecycle controls.
PLO5	Professional Communication	Communicate AI evidence, limitations, recommendations, and tradeoffs to technical and nontechnical stakeholders.
PLO6	Domain Integration	Integrate AI methods with healthcare, business, cybersecurity, or cross-domain professional practice.

Course Learning Outcomes

Outcome	Measurable Course Outcome
CO1	Analyze professional problems in Deep Learning & Neural Networks and formulate AI use cases with explicit stakeholders, decision boundaries, data assumptions, and success criteria.
CO2	Execute or interpret reproducible notebook-based investigations that demonstrate core Deep Learning & Neural Networks methods using guided Python/Colab workflows.
CO3	Evaluate model, workflow, or governance evidence for accuracy, validity, uncertainty, bias, security, privacy, and operational limitations appropriate to core MSAI contexts.
CO4	Produce professional artifacts for Deep Learning & Neural Networks, including briefs, model cards, risk registers, evaluation memos, implementation plans, or executive recommendations.
CO5	Apply responsible AI, academic integrity, data stewardship, accessibility, and human-oversight expectations to course work and proposed deployments.
CO6	Communicate AI findings, limitations, tradeoffs, and next-step recommendations to technical and nontechnical stakeholders using clear graduate-level evidence.

Course-to-Program Outcome Map

Legend: I = introduced, R = reinforced, M = mastered or directly assessed at graduate level.

Course Outcome	PLO1	PLO2	PLO3	PLO4	PLO5	PLO6	Primary Evidence
CO1	M	I	R	I	R	M	Module assignments and scenario framing
CO2	I	M	R	I	R	R	Notebook labs and reproducibility checks
CO3	R	R	M	R	M	M	Rubric risk/limitation criteria and evaluation memos
CO4	R	R	R	M	M	M	Professional artifacts and final portfolio
CO5	R	R	M	M	R	M	Responsible AI, data/privacy, and integrity statements
CO6	R	R	R	R	M	M	Briefs, presentations, reflections, and stakeholder recommendations

Module-Level Assessment Alignment

Module	Topic	Essential Question	Direct Evidence	Outcome Alignment
1	From neurons to multilayer networks	How does a stack of differentiable units approximate complex functions from data?	Module 1 Assignment: Baseline MLP design brief; notebook lab; rubric; reflection	CO1, CO2, CO6
2	Backpropagation and automatic differentiation	How does the chain rule enable efficient learning in deep networks?	Module 2 Assignment: Gradient trace and autograd check; notebook lab; rubric; reflection	CO2, CO3, CO6
3	Optimization, loss, and regularization	How do we train deep networks reliably when loss surfaces are non-convex and data are noisy?	Module 3 Assignment: Optimizer and regularization comparison; notebook lab; rubric; reflection	CO2, CO3, CO6
4	Convolutional neural networks for vision	Why do convolutions, pooling, and translation equivariance matter for image understanding?	Module 4 Assignment: CNN architecture memo; notebook lab; rubric; reflection	CO3, CO4, CO6
5	Sequence models: RNNs and LSTMs	How do recurrent architectures model temporal structure and long-range dependencies?	Module 5 Assignment: Sequence modeling design note; notebook lab; rubric; reflection	CO3, CO5, CO6
6	Attention and transformers	How does self-attention replace fixed recurrence for language and multimodal tasks?	Module 6 Assignment: Attention mechanism analysis; notebook lab; rubric; reflection	CO4, CO5, CO6
7	Generative models and applications	What distinguishes discriminative training from generative modeling?	Module 7 Assignment: Discriminative vs generative comparison; notebook lab; rubric; reflection	CO4, CO5, CO6
8	GPU workflows, scale, and deployment	How do practitioners move from notebook experiments to reproducible GPU training pipelines?	Module 8 Assignment: Reproducible training workflow checklist; notebook lab; rubric; reflection	CO1, CO4, CO6

Direct Assessment Evidence

Direct evidence is gathered from module assignments, notebook lab reflections, rubric scores, mid-course synthesis work, and the final applied portfolio artifact. Each module rubric requires technical or conceptual correctness, evidence use, tradeoff analysis, risk/limitation analysis, and professional communication. The final portfolio artifact should be retained as the primary summative evidence for institutional assessment.

Mastery Thresholds

• Course-level mastery: 80% or higher overall with final portfolio submission.

• Outcome-level mastery: average rubric performance of satisfactory or better on artifacts aligned to the outcome.

• Graduate distinction: evidence is reproducible or auditable, uses domain vocabulary accurately, identifies limitations, and communicates a defensible recommendation.

• Remediation trigger: outcome performance below 80%, missing final portfolio evidence, repeated inability to state limitations, or repeated unsupported claims.

Regular and Substantive Interaction Evidence

The course includes weekly instructor interaction through lecture or narration, guided labs, discussion prompts, office hours or appointments, assignment feedback, and revision guidance. Each module provides slide material sufficient for a 90-minute class session and narration guidance for asynchronous delivery. Students receive structured opportunities to ask questions, compare interpretations, and revise evidence-based claims.

Accessibility and Student Support Evidence

The course uses HTML book pages, downloadable notebooks, Colab launch links, and Codespaces launch links. Colab is positioned as the default first-run environment for students new to Python. Instructors should provide alternative access paths when hardware, account, disability, assistive-technology, or bandwidth constraints interfere with participation. Accessibility and accommodation language appears in the syllabus and should be reconciled with institutional policy before delivery. Program-level support controls, launch checks, and post-offering evidence expectations are documented in CastaliaInstitute/MSAI/accreditation/AINS-accessibility-and-student-support-audit.md.

Academic Integrity and AI-Use Controls

The syllabus permits disclosed AI assistance for brainstorming, debugging, and prose improvement while requiring students to own final reasoning, citations, code behavior, and claims. Rubrics and reflections require students to explain their work, name assumptions, and document limitations, reducing the risk of unexamined AI-generated submissions.

Continuous Improvement Plan

At the end of each offering, the program should review: grade distributions, rubric outcome performance, final portfolio samples, student feedback, instructor notes, lab completion friction, accessibility concerns, and evidence of outcome mastery. At least one improvement action should be documented for the next term. Recommended evidence retention includes anonymized high/mid/low artifacts, rubric summaries, syllabus version, assignment prompts, and instructor reflection.

Assessment Evidence Template

The companion assessment-evidence.md file provides the post-offering evidence-retention template for rubric summaries, anonymized artifact samples, calibration records, and continuous-improvement actions. This file should be completed after each course offering and retained with institution-controlled assessment records.

Program Assessment Cycle Evidence

Course assessment evidence rolls up into the AINS program assessment cycle documented in CastaliaInstitute/MSAI/accreditation/AINS-program-assessment-cycle.md. The program cycle defines the PLO curriculum map, annual assessment calendar, benchmark thresholds, evidence sampling rules, and closing-the-loop expectations. Course-level rubric summaries and artifact samples should be retained using assessment-evidence.md and reviewed against those program benchmarks after each offering.

Scholarly Readings and Standards Evidence

The syllabus includes module-level required readings, professional standards, or applied resources. Program-level source coverage, access-verification expectations, and standards-currency checks are documented in CastaliaInstitute/MSAI/accreditation/AINS-scholarship-and-standards-audit.md. Before each offering, the instructor or program designee should verify library access, open-access links, edition currency, accessibility, and any instructor-selected substitutions.

Institutional Policy Reconciliation

Course-level accreditation evidence should be reviewed alongside the central Aurnova policy reconciliation crosswalk in CastaliaInstitute/MSAI/accreditation/AINS-policy-reconciliation-crosswalk.md and the final readiness audit in CastaliaInstitute/MSAI/accreditation/AINS-final-accreditation-readiness-audit.md. Public syllabus language is accreditation-ready for curriculum review, but it must be reconciled with official Aurnova catalog, handbook, accessibility, registrar, LMS, faculty credential, privacy, and compliance records before delivery or formal submission.

Accreditation Review Notes

This course is strongest when reviewed as applied graduate professional education rather than as a computer science programming course. The instructional design emphasizes AI literacy, evidence interpretation, responsible use, domain judgment, and stakeholder communication while still providing guided technical exposure through notebooks.