Skip to content

curriculum(SEF8): competency-derived numerical mathematics and inference progression - #164

Merged
christophersesugh merged 1 commit into
mainfrom
curriculum/sef8-competency-derived-numerical-mathematics
Nov 18, 2026
Merged

christophersesugh merged 1 commit into
mainfrom
curriculum/sef8-competency-derived-numerical-mathematics

Conversation

@christophersesugh

Copy link
Copy Markdown
Collaborator

M1 — SEF8 mathematics for numerical computing & AI: competency-derived reassessment

Pinned authoritative baseline: main at c0a01421b4c95bc661bd785121e3b737d7964f42, after approved SEF5 PR #162.
Repeatable review protocol: temporary v1.0 review guidelines, intentionally unmerged PR #150, tracked from M0 inventory #152. Binding OEE spec > design guide > accepted course/catalog decisions.
Files: curriculum/SEF/SEF8.md, tests/test_sef8_migration.py. No additional learner assessment, normative rules, permanent audits, cross-course changes or modifications to separately ongoing AI-program branches. This review PR targets main directly.

1. Independently justified exit capability and scope

A SEF11-qualified learner can derive and defend finite numerical mathematics in an actual Python 3.11+ standard-library setting, not merely memorize formulas or trust library output: establish function domains and calculus limits, manipulate vectors/matrices and classify linear systems, calculate finite probabilities and responsible descriptive/statistical inference, derive analytic rates and continuous accumulation, distinguish exact identities from floating-point approximations, analyze finite-difference error and bounded gradient descent, and reproduce each claim with independent oracles, clear invalid-input policies and an unfamiliar-case defense.

Evidence-based external comparisons (disciplinary references, not prescriptions of module count or proof of OEE accreditation):

Prerequisite boundary: SEF11 teaches arithmetic, elementary algebra and functions, finite probability counting prerequisites, proof, recurrences and discrete graphs; SEF8 uses this math to develop calculus, linear algebra, statistical inference and numerical precision. It does not add an admissions exam or require completion of the concurrently available SEF4 algorithms course. Advanced statistical modeling and production AI training belong in separately governed later programs; these are foundation-level, mathematically honest examples.

2. All outcomes and assessments reviewed; preservation choice

The current fifteen published learning outcomes are SEF8-O19, O3, O6, O7, O8, O9, O10, O11, O12, O14, O15, O16, O17, O18 and O20. Their identifiers are historical, not a numbering template. The whole Learning outcomes section is byte-for-byte identical to the pinned baseline.

The six summative assessments are SEF8-AS3, AS4, AS6, AS9, AS7 and AS10, intentionally nonconsecutive and in instructional prerequisite order rather than numerical order. The entire Work/Criteria/Evidence section for all six is byte-for-byte identical to the pinned baseline. No extra learner exam, coding package, notebook project or hidden grading mechanism is introduced. The teaching is designed to make these existing contracts legitimately learnable and independently checkable.

3. Complete module/lesson reassessment and justified sequence

Original 8 modules, 59 numbered lessons showed genuine duplicated foundations:

  • Former M2/M3 separately repeated vector spaces, basis/independence, rank/null space and matrix composition; only the numerical-error/eigenpair context introduced substantially different teaching, which belongs after the prerequisite linear transformations/system-solving mechanisms.
  • Former M4/M5 both introduced sampling variability, descriptive statistics and confidence intervals; inferential claims need the preceding probability/variance model, but do not need a second introductory statistics survey.
  • Former M6/M7 repeatedly introduced numerical derivatives and error trade-offs while advanced float/cancellation analysis followed the first numerical examples. Derive analytic gradients and integrals first, then teach finite-precision representations, forward/central approximations and gradient descent as a coherent numerical computation competency.
  • Multiple lessons had only non-observable titles (Vectors, Matrix multiplication, Sampling, bias and variability, Gradient descent and learning-rate behavior). The new descriptions include input/domain assumptions, independent checks and failure/negative cases.

New learning sequence: 7 modules, 55 lessons (7 / 10 / 8 / 7 / 8 / 9 / 6). This is not a target or intentionally varied pattern; module boundaries reflect independently necessary mathematics and safe prerequisites.

New module Distinct intellectual ownership and mandatory knowledge preparation
M1 — Functions, Domains and Limits (7 lessons) Numerical problem/model framing; rational domain exclusions; composition and algebra; exponential/log restrictions; secants; valid algebraic limits; Fraction/math exact-versus-approximate checks. SEF11 algebra is developed into the function/limit foundation, not reexamined as an entry gate.
M2 — Linear Algebra, Systems and Transformations (10) Coordinate vectors → Euclidean norms/dot/orthogonality → shape-valid matrix products → transformation composition → elimination → independence and basis → rank/null-space witnesses → consistent versus inconsistent rank-deficient systems → elementary diagonal eigenpair → independent exact/float checks. Unifies former M2/M3 and covers every small-vector/matrix/system requirement of AS3.
M3 — Probability, Random Variables and Descriptive Statistics (8) Sample spaces/weights → conditional probability including zero denominator → actually checked independence → base-rate Bayes reasoning → discrete distributions → expectation/variance → population vs sample variance and empty cases → seeded two-stage conditional experiment. Supports AS4 without prematurely introducing inference conclusions.
M4 — Sampling Uncertainty and Responsible Inference (7) Estimand and biased sampling → correct Bernoulli plug-in standard error → explicitly approximate normal 95% interval using a supplied 1.96 value → repeated coverage and boundary failures → meaning of a p-value under a null → association/confounding/causal limits → reproducible inference report. No implicit prerequisite that the learner already integrated the normal density in calculus.
M5 — Analytic Calculus and Continuous Accumulation (8) Derivative as limit → elementary rules → chain rule → partial derivatives/gradients → local directional change → Riemann sums → antiderivative/Fundamental Theorem → probability density and interval integral. Does not require a numerical differentiation algorithm before the underlying derivative is defined.
M6 — Floating-Point Reliability, Numerical Gradients and Optimization (9) Binary representability → cancellation and tolerance → forward difference → central difference and truncation assumptions → decreasing-h roundoff trade-off → partial numerical gradients → explicit gradient-descent recurrence → stable/oscillating/divergent step cases → bounded stopping and nonconvergence status. Makes O12/O15/O16 independently observable, with no fictitious guarantee of global optimization.
M7 — Integrated Numerical Work and Independent Defense (6) Reuse all assessed calculations/code; independent analytic and computational checks; invalid-input matrix/probability/numerical cases; actual seeded uncertainty evidence; clean-start accessible notebook/package; genuine unfamiliar reviewer challenge and repair. This is the already published AS10 integration, not a second project.

The six existing assessments remain a course-level criterion, not six mandatory module checklists. Optional practices are formative and not pass gates.

4. Assessment evidence: genuine mathematical claims and negative cases

Existing gate Instruction and validity map Correct independent reasoning and failure controls
AS3 — vectors/matrices/systems (O6/O7/O3/O18) M2 Norm of (3,4) = 5; dot of (3,4) and (-4,3) = 0; [[1,2],[3,4]]·(2,-1) = (0,2); singular [[1,2],[2,4]] has nonzero null vector; uniquely solvable vs inconsistent vs nonunique; shapes, transformation order and simple eigenpair are inspectable
AS4 — Bayes/statistics (O8/O9/O14/O3) M3 Posterior with prevalence .1, sensitivity .8 and false-positive .2 is 0.08/(0.08+0.18)=4/13; probability-zero conditioning is undefined; [2,4,6] mean 4, population variance 8/3, sample variance 4; seeded trials are compared, not forced to equal analytic rates
AS6 — inference (O9/O10/O14/O18) M3→M4 With 60/100, p_hat=0.6, plug-in SE sqrt(.6×.4/100)≈.049 and approximate Wald interval 0.6±1.96·SE≈[.504,.696]; coverage across a real seeded batch estimates rather than guarantees nominal frequency; real bias and p-value/causal limitations explained
AS9 — calculus/limits (O19/O20/O11/O12/O3/O18) M1→M5→M6 Derive f'(2)=7 for f(x)=x²+3x; gradient (8,3) for x²+3xy at (1,2); integral ∫₀¹ 2x dx=1; invalid domain/step rejected; finite-difference tables do not prove calculus results
AS7 — floating point/gradient descent (O11/O12/O15/O16/O18) M5→M6 For f(x)=(x−3)², rate .25 from zero yields x1=1.5, x2=2.25; rate 1.0 yields the 0↔6 oscillation, 1.1 diverges; reject invalid/nonfinite rates, require bounded attempts and nonconvergence report
AS10 — integration/defense (all 15 outcomes) M7 and prior evidence The same standard-library notebook/small library reuses independent expected values, seed and tolerance records, real error fixtures and an unfamiliar review variation; no new hidden capstone

These concrete arithmetic checks are reproducible mathematical cross-checks, not a claim of actual learner assessment.

5. Dispositions, risks and unresolved calibration

  • F1 — major: duplicated linear-algebra foundations. M2/M3 both taught independence, rank, composition. Resolved: one coherent progression from vectors to system classification, eigen witness and exact/approximate checks.
  • F2 — major: inferential concepts taught before distinct modeling prerequisites. Old M4/M5 repeated sampling and intervals. Resolved: basic probability and descriptive statistics → explicit inferential estimates/coverage/bias, with a supplied normal critical value rather than covert calculus requirements.
  • F3 — major: numerical derivatives and float error separated and repeated. Old M6/M7 repeated gradient/finite differences without clearly distinguishing analytic derivation, truncation and rounding. Resolved: analytic calculus M5 then numerical reliability/gradients/descent M6.
  • F4 — major: lesson titles not demonstrable. Noun-style introductory lessons did not require domain, counterexample, error or independent check. Resolved: each lesson names one primary mathematical method and inspectable mechanism/failure.
  • F5 — major: fixed editorial quantity in test suite. tests/test_sef8_migration.py required exactly eight modules. Resolved: sequential module IDs and meaningful lesson structure plus genuine prerequisite ordering (probability before inferential, calculus before numeric differences and descent, all core methods before final integration), while retaining nonconsecutive historical assessment IDs.
  • F6 — minor: coverage/roundoff claims easily overstated. A Wald interval can be poor for near-zero proportions and tiny samples, and a shrinking finite-difference h can worsen roundoff/cancellation. Resolved in teaching: explicit limits, negative cases and actual finite evidence without altering published completion criteria.
  • F7 — not ready for a readiness promotion: 12–14 weeks at provisional 10–15 h/wk is not independently calibrated for learners from SEF11. Actual attempts, accessible tools, failures and reviewer time must inform later revision. No change from draft.
  • F8 — potential downstream content synchronization: Separately developed AI-program material should be checked after its independent PRs merge, to ensure it refers to the actual SEF8 math preparation rather than the pre-review module numbers. No unmerged AI content was changed in this PR.

6. Exact verification, DCO and merging procedure

Source checks on the branch:

  • Learning outcomes and all Required summative assessments Work/Criteria/Evidence sections remain byte-for-byte identical to main at baseline.
  • All 15 outcomes have real module ownership; seven modules with sequential IDs and 55 explicitly described lessons; six nonconsecutive historical assessment identifiers retained.
  • SEF11 prerequisite, draft label, OEE v0.2.0, duration and Python standard-library-only requirements unchanged.
  • tests/test_sef8_migration.py now checks structural competency order, detailed lesson objectives, key mathematical safety mechanisms and assessment identity without a fixed module count.
  • No new normative guide, admission prerequisite, external scientific dependency, mandatory learning activity or permanent audit document.

Still required: review repository CI, personally sign the commits to satisfy DCO and await maintainer approval. Passing CI validates repository structure and tests, not real academic learner outcomes.

The GitHub API-generated commits on this branch lack a real person's DCO attestation. In a clean local OEE checkout, the maintainer can squash them into a single personally attested commit:

git fetch origin
git switch curriculum/sef8-competency-derived-numerical-mathematics
git pull --ff-only origin curriculum/sef8-competency-derived-numerical-mathematics
git reset --soft c0a01421b4c95bc661bd785121e3b737d7964f42
git commit -s -m "curriculum(SEF8): competency-derived numerical mathematics progression"
git push --force-with-lease origin HEAD

Do not merge until required checks and the maintainer's review pass. Protocol PR #150 remains unmerged; this PR targets main directly. Next M1 review after approval/merge: SEF6 — C Programming and Memory.

Signed-off-by: Christopher S. Aondona <christohybrid185@gmail.com>
@christophersesugh
christophersesugh force-pushed the curriculum/sef8-competency-derived-numerical-mathematics branch from f3ea0fc to 96141e4 Compare November 18, 2026 12:59
@christophersesugh
christophersesugh merged commit 828a12a into main Nov 18, 2026
7 checks passed
@christophersesugh
christophersesugh deleted the curriculum/sef8-competency-derived-numerical-mathematics branch November 18, 2026 13:00
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant