Fig. 4: Sensitivity of the machine learning assisted vector magnetometer.

a Neural network predicted sensitivity for field amplitude (a1) and orientations (a2, a3) at low frequency. The measurement is performed at θ = 63.435°, φ = 60° for a field magnitude about 140 nT. b NN-predicted change of the magnetic field magnitude (on top of 140 nT, b1) and directions (b2, b3) versus the corresponding true values. The dashed line corresponds to the y = x function. The results are demonstrated for magnetic field changes at a frequency of 11 Hz. The size of the error bars (standard deviation from 60 repetitive independent measurements) are in agreement with the NN predicted sensitivity level at 11 Hz.