zjs81/meshcore-open
1
0
Fork 0
mirror of https://github.com/zjs81/meshcore-open.git synced 2026-06-21 01:45:47 +10:00
Code Issues Packages Projects Releases 2 Wiki Activity

fix: address PR #296 code review feedback

Browse Source 
- Clamp ML predictions between physics floor (raw airtime) and ceiling
  (worst-case formula) so model can never produce unsafe timeouts
- Replace hourOfDay feature with secondsSinceLastRx for network activity
- Remove unused _ContactStats.stdDev and dead model persistence code
- Debounce observation writes (2s) instead of writing on every delivery
- Skip recording observations when pathLength is null to avoid corrupting
  training data
- Add comment explaining global (not per-contact) RX time tracking
- Remove notifyListeners from retrain to avoid unnecessary widget rebuilds
- Run dart format
This commit is contained in:
zjs81
2026-03-14 17:32:08 -07:00
parent 2ee2358ecc
commit b336aedbc5
6 changed files with 187 additions and 142 deletions
Download Patch File Download Diff File Expand all files Collapse all files
test/services/ml_algo_sanity_test.dart
+85 -51
View File
@@ -6,18 +6,22 @@ import 'package:ml_dataframe/ml_dataframe.dart';
void main() {
test('LinearRegressor basic sanity check', () {
// Simple: y = 2x + 100
final data = DataFrame([
[1.0, 102.0],
[2.0, 104.0],
[3.0, 106.0],
[4.0, 108.0],
[5.0, 110.0],
[10.0, 120.0],
[20.0, 140.0],
[50.0, 200.0],
[0.0, 100.0],
[100.0, 300.0],
], headerExists: false, header: ['x', 'y']);
final data = DataFrame(
[
[1.0, 102.0],
[2.0, 104.0],
[3.0, 106.0],
[4.0, 108.0],
[5.0, 110.0],
[10.0, 120.0],
[20.0, 140.0],
[50.0, 200.0],
[0.0, 100.0],
[100.0, 300.0],
],
headerExists: false,
header: ['x', 'y'],
);
debugPrint('Training data columns: ${data.header}');
debugPrint('Training data rows: ${data.rows.length}');
@@ -25,7 +29,9 @@ void main() {
final model = LinearRegressor(data, 'y');
final testDf = DataFrame(
[[25.0]],
[
[25.0],
],
headerExists: false,
header: ['x'],
);
@@ -38,45 +44,63 @@ void main() {
test('LinearRegressor multi-feature with constant column produces zeros', () {
// isFlood=0 for all rows → zero-variance column → singular matrix
final data = DataFrame([
[0.0, 50.0, 14.0, 0.0, 1900.0],
[0.0, 80.0, 14.0, 0.0, 2200.0],
[2.0, 50.0, 14.0, 0.0, 5000.0],
[4.0, 50.0, 14.0, 0.0, 9500.0],
], headerExists: false, header: [
'pathLength', 'messageBytes', 'hourOfDay', 'isFlood', 'deliveryMs',
]);
final data = DataFrame(
[
[0.0, 50.0, 14.0, 0.0, 1900.0],
[0.0, 80.0, 14.0, 0.0, 2200.0],
[2.0, 50.0, 14.0, 0.0, 5000.0],
[4.0, 50.0, 14.0, 0.0, 9500.0],
],
headerExists: false,
header: [
'pathLength',
'messageBytes',
'hourOfDay',
'isFlood',
'deliveryMs',
],
);
final model = LinearRegressor(data, 'deliveryMs');
final testDf = DataFrame(
[[2.0, 50.0, 14.0, 0.0]],
[
[2.0, 50.0, 14.0, 0.0],
],
headerExists: false,
header: ['pathLength', 'messageBytes', 'hourOfDay', 'isFlood'],
);
final pred = model.predict(testDf).rows.first.first;
debugPrint('With constant isFlood column: hops=2 → ${(pred as num).round()}ms (likely 0)');
debugPrint(
'With constant isFlood column: hops=2 → ${(pred as num).round()}ms (likely 0)',
);
});
test('LinearRegressor 2-feature works correctly', () {
// Just pathLength + messageBytes → deliveryMs
final data = DataFrame([
[0.0, 50.0, 1900.0],
[0.0, 80.0, 2200.0],
[2.0, 50.0, 5000.0],
[2.0, 80.0, 5500.0],
[4.0, 50.0, 9500.0],
[4.0, 80.0, 10000.0],
[0.0, 30.0, 1800.0],
[2.0, 30.0, 4800.0],
[4.0, 30.0, 9000.0],
[0.0, 60.0, 2000.0],
], headerExists: false, header: ['pathLength', 'messageBytes', 'deliveryMs']);
final data = DataFrame(
[
[0.0, 50.0, 1900.0],
[0.0, 80.0, 2200.0],
[2.0, 50.0, 5000.0],
[2.0, 80.0, 5500.0],
[4.0, 50.0, 9500.0],
[4.0, 80.0, 10000.0],
[0.0, 30.0, 1800.0],
[2.0, 30.0, 4800.0],
[4.0, 30.0, 9000.0],
[0.0, 60.0, 2000.0],
],
headerExists: false,
header: ['pathLength', 'messageBytes', 'deliveryMs'],
);
final model = LinearRegressor(data, 'deliveryMs');
for (final hops in [0.0, 2.0, 4.0]) {
final testDf = DataFrame(
[[hops, 50.0]],
[
[hops, 50.0],
],
headerExists: false,
header: ['pathLength', 'messageBytes'],
);
@@ -87,20 +111,28 @@ void main() {
test('LinearRegressor multi-feature with variance in all columns', () {
// Mix flood and direct so isFlood has variance
final data = DataFrame([
[0.0, 50.0, 14.0, 0.0, 1900.0],
[0.0, 80.0, 10.0, 0.0, 2200.0],
[2.0, 50.0, 16.0, 0.0, 5000.0],
[2.0, 80.0, 20.0, 0.0, 5500.0],
[4.0, 50.0, 8.0, 0.0, 9500.0],
[4.0, 80.0, 12.0, 0.0, 10000.0],
[-1.0, 40.0, 14.0, 1.0, 5000.0],
[-1.0, 60.0, 18.0, 1.0, 6500.0],
[-1.0, 30.0, 10.0, 1.0, 4000.0],
[-1.0, 80.0, 22.0, 1.0, 7000.0],
], headerExists: false, header: [
'pathLength', 'messageBytes', 'hourOfDay', 'isFlood', 'deliveryMs',
]);
final data = DataFrame(
[
[0.0, 50.0, 14.0, 0.0, 1900.0],
[0.0, 80.0, 10.0, 0.0, 2200.0],
[2.0, 50.0, 16.0, 0.0, 5000.0],
[2.0, 80.0, 20.0, 0.0, 5500.0],
[4.0, 50.0, 8.0, 0.0, 9500.0],
[4.0, 80.0, 12.0, 0.0, 10000.0],
[-1.0, 40.0, 14.0, 1.0, 5000.0],
[-1.0, 60.0, 18.0, 1.0, 6500.0],
[-1.0, 30.0, 10.0, 1.0, 4000.0],
[-1.0, 80.0, 22.0, 1.0, 7000.0],
],
headerExists: false,
header: [
'pathLength',
'messageBytes',
'hourOfDay',
'isFlood',
'deliveryMs',
],
);
final model = LinearRegressor(data, 'deliveryMs');
@@ -116,7 +148,9 @@ void main() {
header: ['pathLength', 'messageBytes', 'hourOfDay', 'isFlood'],
);
final pred = model.predict(testDf).rows.first.first;
debugPrint('4-feature: hops=${tc[0]} flood=${tc[3]} → ${(pred as num).round()}ms');
debugPrint(
'4-feature: hops=${tc[0]} flood=${tc[3]} → ${(pred as num).round()}ms',
);
}
});
}
test/services/timeout_prediction_service_test.dart
+3 -3
View File
@@ -64,9 +64,9 @@ void main() {
expect(direct4!, greaterThan(direct2!));
expect(direct2, greaterThan(direct0!));
// All should be within the clamp range
expect(direct0, greaterThanOrEqualTo(2000));
expect(direct4, lessThanOrEqualTo(120000));
// All should be positive
expect(direct0, greaterThan(0));
expect(direct4, greaterThan(0));
// Print predictions for visibility
debugPrint('Predictions (with 1.5x safety margin):');