Closed Joshuaalbert closed 4 months ago
Only first few real components are changing. This probably implies some index is being selected inadvertently. Also, explains why residuals are not super tight.
"gains": { "type": "numpy.ndarray", "array_real": [ [ [ [ [ [ 0.9810193777084351, -0.1584736406803131 ], [ 0.11337465047836304, 1.0116183757781982 ] ], [ [ 1.024043321609497, 0.004812115803360939 ], [ 0.112271748483181, 0.9224221706390381 ] ], [ [ 1.0020111799240112, -0.26801177859306335 ], [ 0.3347955048084259, 1.0057669878005981 ] ] ], [ [ [ 1.0, 0.0 ], [ 0.0, 1.0 ] ], [ [ 1.0, 0.0 ], [ 0.0, 1.0 ] ], [ [ 1.0, 0.0 ], [ 0.0, 1.0 ] ] ], [ [ [ 1.0, 0.0 ], [ 0.0, 1.0 ] ], [ [ 1.0, 0.0 ], [ 0.0, 1.0 ] ], [ [ 1.0, 0.0 ], [ 0.0, 1.0 ] ] ], [ [ [ 1.0, 0.0 ],
I think what's happening is predict code is broadcasting somewhere where it shouldn't. Hence solution is a unittest for all predict implentations.
Seems that only first antenna is corrected
Only first few real components are changing. This probably implies some index is being selected inadvertently. Also, explains why residuals are not super tight.