Linked ROM / FlashKB
Not measuredAdvisor / Committee Review
Evidence-driven review brief
This review brief summarizes the project's current evidence, unresolved scientific decisions, and immediate execution constraints for advisor and committee discussion. It distinguishes completed experiments, unsupported candidate claims, retained failed outcomes, frozen protocols, blocked hardware work, and proposed contributions.
Research problem and provisional gap
Electronic-nose sensors drift through aging, contamination, and environmental variation. Chronological evaluation tests future-batch degradation directly; random splits can mix future characteristics into training and give an overly favourable deployment estimate.
The literature reviewed so far has not identified a study that unifies chronological e-nose drift evaluation, quantitatively assessed explanation fidelity and stability, and physically measured TinyML deployment cost in one traceable evidence chain. This remains a provisional literature finding rather than a claim of exhaustive novelty.
Research design
FIXED_ORIGIN is the primary prospective protocol; EXPANDING_WINDOW is secondary adaptation analysis; IID_DIAGNOSTIC is diagnostic only. Preprocessing is training-only and four frozen model families underpin the current results. configs/chronological_protocol.yaml
Current evidence
Executed 16Protocol frozen 2Failed 2Blocked 6Running 1
Counts are derived from the declared pipeline registry, not a progress percentage. Negative outcomes remain distinct from unexecuted work.
| Area | Status | Boundary |
|---|---|---|
| 10 — Fidelity evaluation | Executed | EXP-XAI-FIDELITY-001; method-specific global, local, sensor-group, family, matched-random, and bootstrap evidence. Local ablation is labelled ABLATION_CONSISTENCY because its perturbation evaluation is partly circular. |
| 11 — Explanation stability evaluation | Executed | EXP-XAI-STABILITY-001; global chronological, physical sensor/family, natural-neighbor, matched cross-sectional, relative, and fidelity-linkage evidence. No universal stability score. |
| 12 — Controlled host-side XAI computational cost | Executed | EXP-XAI-LATENCY-001 measured controlled single-thread workstation cost with raw nanosecond timings and hardware-independent operation counts. Global and local scopes remain separate. Host latency is not an estimate of nRF52840 latency or energy; MCU latency, physical energy, Flash, and SRAM remain NOT_MEASURED/BLOCKED. Stage 13 did not execute. |
| 13 — Embedded architecture and export protocol freeze | Protocol frozen | GATE-EMBED-EXPORT-001 architecture and numerical-equivalence protocol frozen. This is not an executed model export or performance experiment. Export, quantization, firmware, compiled Flash/SRAM, MCU timing, and energy remain NOT_EXECUTED/NOT_MEASURED. |
| 14 — Standalone FP32 export equivalence | Failed | EXP-EMBED-FP32-EQUIV-001 completed with scientific outcome FAILED. C1 and C4 passed score, probability, normalization, golden decision, and boundary decision checks but failed frozen preprocessing criteria. Host-compiled validation only; quantization and MCU work remain unexecuted. |
| 14R — C1 explicit FP32 preprocessing repair | Failed | EXP-EMBED-C1-PREPROC-REPAIR-001 completed with outcome FAILED. P0 baseline reproduced; P1-P4 all failed unchanged universal preprocessing criteria. C4 remained frozen failed, and quantization, firmware, MCU measurement, and hardware resource claims remain unexecuted. |
| 15 — nRF52840 deployment | Hardware blocked | EXP-MCU-C1-FP32-PORT-001 is BLOCKED_HARDWARE. No physical Nordic/J-Link/CMSIS-DAP device was detected, so no target was guessed, no toolchain was changed, and no build, flash, physical correctness, linked ROM, or linked static-RAM evidence was produced. |
Findings requiring interpretation
Stage 10 fidelity and Stage 11 stability experiments were executed, but their tested candidate claims were unsupported. These retained results may indicate weak explanation behaviour, limitations in the tested definitions, or both; this branch does not alter metrics or rerun experiments. results/xai/stage10_fidelity_summary.csv results/xai/stage11_stability_summary.csv
Stages 14 and 14R retain failed export/preprocessing outcomes; later host equivalence does not establish physical deployment.
Hardware constraint
Prerequisite detection was executed, but no supported nRF52840 board or debug interface was detected. Board identity was not guessed. Physical deployment, Flash/SRAM, MCU latency, and PPK2 energy remain unmeasured.
PLANNED CONNECTION — NOT PHYSICALLY EXECUTED
- 1Model exportHost artifact available
- 2Firmware buildBlocked by hardware
- 3Board programmingBlocked by hardware
- 4Functional validationBlocked by hardware
- 5Repeated inferenceBlocked by hardware
- 6PPK2 trace captureBlocked by hardware
- 7Trigger-window extractionBlocked by hardware
- 8Energy integrationBlocked by hardware
- 9Evidence registrationBlocked by hardware
E = ∫ V(t)I(t)dt. No post-detection step is shown as completed.| Measurement | Unit | Current state |
|---|---|---|
| Linked ROM / Flash | KB | Not measured |
| Static / peak SRAM | KB | Not measured |
| MCU inference latency | ms | Not measured |
| MCU explanation latency | ms | Not measured |
| PPK2 inference energy | µJ | Not measured |
| PPK2 explanation energy | µJ | Not measured |
Static / peak SRAMKB
Not measuredMCU inference latencyms
Not measuredMCU explanation latencyms
Not measuredPPK2 inference energyµJ
Not measuredPPK2 explanation energyµJ
Not measuredHardware record · TinyML record · results/embedded/stage15_hardware_detection.json
Proposed systems contribution
A reproducible systems-level study connecting chronological drift evaluation, resource-aware explanation analysis, numerical equivalence, and physical deployment cost. It is not necessarily a new learning algorithm, and final scope depends on remaining hardware work.
Discussion priorities
Evaluation hierarchy and model scope
Whether FIXED_ORIGIN should remain primary, EXPANDING_WINDOW secondary, IID_DIAGNOSTIC diagnostic only, and whether a compact deep model is scientifically necessary.
Interpretation of explanation evidence
Stages 10–11 were executed but their tested candidate claims were unsupported. Guidance is requested on whether this reflects weak explanation behaviour, definitions requiring refinement, or a need for sensitivity analysis.
Publication evidence bar
Advice is requested on venue positioning, physical nRF52840/PPK2 requirements, an additional e-nose dataset, and whether a systems contribution is sufficient without a new algorithm.
Evidence and collaboration notices
Evidence last exported 2026-08-14T21:21:27Z from commit d4c1b0b37dd4 on main. Reviewer uploads and comments are collaborative material, not validated scientific evidence, unless separately reviewed and registered through the evidence pipeline.
Discussion
Published discussion will appear here. Reviewer comments represent contributor perspectives, not verified project findings; they become evidence only through the separate evidence pipeline.