Fig. 4.

Probability of agreement on synthetic data. Classification performance on synthetic data of varying data density $\overline{\left|\mathrm{\Omega }\right|}$. Model performance is given as the probability of agreement [13] score (${\mathrm{\Phi }}_s$ from (8)) with $95\%$ CI. Higher ${\mathrm{\Phi }}_s\in [0,1]$ signifies better model fit. The prediction models are Matrix Factorization (MF), Convolutional MF (CMF) and Shifted CMF (SCMF)