New technical note AGE-PN-001: reproducible benchmark for WDM stability See evidence
Free evaluation

One log. One metric. One useful decision.

The free demonstration evaluates, offline, whether shared structure in your data can outperform an agreed baseline. It requires neither chip access nor control integration.

Closed scope

The free phase answers one question.

Does a cross-channel estimator consistently reduce error against the selected baseline on a segment that was not used for fitting?

Included

Minimum preparation

Review of schema, units, cadence, channels and basic log quality.

Included

Offline comparison

Baseline and structured estimator under the same window, metric and rules.

Included

Decision note

Quantitative result, uncertainty, observed limitations and go / no-go recommendation.

—

It does not include real-time integration, actuator control, an experimental campaign, full software adaptation or a guarantee of operational improvement.

Minimum input

Multichannel temporal data with documented physical meaning.

No proprietary format is required. The open package includes an example CSV and a specification that can be mapped to HDF5, Parquet, MAT or other formats.

  • Monotonic timestamp
  • Channel or device identifier
  • Detuning, wavelength or equivalent error
  • Units, cadence and missing-value treatment
  • Actuator state when it affects observation
minimal_schema.csvexample
timestamp_s,channel_id,detuning_pm,heater_cmd
0.0,       ring_01,   +2.14,       0.312
60.0,      ring_02,   +1.87,       0.296
120.0,     ring_03,   +2.41,       0.301
...

# units and acquisition schedule required
Evaluation protocol

The prediction is fixed before the held-out segment is scored.

The package includes a SHA-256 sealed workflow to record a prediction before scoring the reference. It is a traceability aid, not a substitute for independent audit or a complete cryptographic guarantee.

See the full protocol →

01agree

Fix

Schema, metric, baseline, temporal split and failure conditions.

02identify

Fit

Estimate the model using only the allowed identification window.

03commit

Seal

Generate the prediction and record its hash with the run manifest.

04score

Evaluate

Compare with held-out truth and report metrics, failures and sensitivity.

Short reportdelivery
Log quality and coverageyes
Baseline erroryes
Structured-estimator erroryes
Interval / sensitivityyes
Recommendation and limitsyes
Interpretation

Three possible outcomes. All three are useful.

  • Positive signal: held-out improvement supports a platform diagnostic.
  • Inconclusive result: more duration, excitation or quality is needed to separate dynamics.
  • Negative signal: no robust advantage is seen, so integration is not recommended.
Next step

Do not send data yet. Describe the log first.

Platform, channel count, duration, cadence, measured variable and current baseline are enough to assess technical fit.