Fig. 2: Estimation of the optimum number of clusters (Kopt). | Nature Communications

Fig. 2: Estimation of the optimum number of clusters (Kopt).

From: Machine learning for microscopy data analytics targeting real-time optical characterization of semiconductor nanocrystals

Fig. 2: Estimation of the optimum number of clusters (Kopt).

a–d The <SS > , <%Mscl > , <CH index > , and <SSD> metric profiles, considering 100 iterations of the code, are calculated for both unprocessed and the binned (by frames) PL traces following data normalization. Metric profiles are calculated for the data binned by 10, 20, 25, 40, 50, 80, 100, 125, 200, 250, 400, and 500 frames and compared with that for the pristine (raw) blinking trajectories (in black). The ideal window of binning is estimated at 80 frames, with the corresponding average metric profiles represented in cyan, as highlighted in each panel. The insets display zoomed view of the average metric profiles with the Kopt indicated therein. Average metric profiles for the raw traces (in black) and optimally binned traces (binned by 80 frames, in cyan) are kept in bold.

Back to article page