Figure 4—figure supplement 1.

Figure 4—figure supplement 1. Smoothing the calcium imaging signal can improve the accuracy of GC.(A) We smooth the noisy calcium imaging signal f = DF/F using total-variational regularization (see Materials and Methods), and plot the example traces of the original and the maootherned neuronal signals, and the residual noise, using the dataset f3t2(N = 11 neurons) from De Vico Fallani et al., 2014. (B) The Pearson correlation coeffecient show the residual noise is correlated, (C) GC networds constructed using the original noisy calcium signal compared to ones using the smoothed signal for motorneurons. (D) Weight of ipsilateral GC links, W_IC and the weight of rostal-to-cordal links, W_RC, for the original calcium signals and smoothened signals. We expect W_IC ~ 1 and a null model with no bias for where the links are placed will have W_IC = 0.5. The x-axis is sorted using the W_IC value for bivariate GC analysis with the original calcium data. Data points are connected with lines to guide the eye. The bivariate GC results are plotted with purple circles, and the multivariate GC results with green diamond. GC analysis results using the original noisy fluroscence signals are shown with empty markers, while the results from first smoothed data are shown with filled markers.