Table 5.

10-fold cross-validated test data log-likelihood of the proposed method Vs vanilla LDA.

Test data log-likelihood of the proposed method
Model hyperparameters	Test log-likelihood for the model as in33 (mean ± std)	Test log-likelihood for the proposed model (mean ± std)
K = 2, α = 0.1, β = 0.1	−687951.86 (±47455.33)	−367855.70 (±27324.98)
K = 2, α = 0.1, β = 0.01	−688027.21 (±47424.77)	−367728.17 (±27179.73)
K = 2, α = 0.1, β = 0.001	−688049.86 (±47477.81)	−367781.96 (±27190.10)
K = 2, α = 0.01, β = 0.1	−689056.69 (±47443.14)	−368086.97 (±27197.90)
K = 2, α = 0.01, β = 0.01	−689270.81 (±47284.88)	−368368.67 (±27174.22)
K = 2, α = 0.01, β = 0.001	−689730.70 (±47553.67)	−368588.16 (±27689.93)
K = 2, α = 0.001, β = 0.1	−693364.63 (±47505.99)	−370060.29 (±27350.08)
K = 2, α = 0.001, β = 0.01	−693446.41 (±47418.04)	−369991.03 (±27165.18)
K = 2, α = 0.001, β = 0.001	−693645.14 (±47480.30)	−370052.10 (±27195.04)
K = 3, α = 0.1, β = 0.1	−681064.83 (±47169.90)	−364978.13 (±27062.30)
K = 3, α = 0.1, β = 0.01	−681003.93 (±47332.34)	−365462.25 (±27289.95)
K = 3, α = 0.1, β = 0.001	−681534.16 (±47021.27)	−365380.51 (±27122.67)
K = 3, α = 0.01, β = 0.1	−682631.13 (±47218.50)	−365766.74 (±26798.13)
K = 3, α = 0.01, β = 0.01	−682392.99 (±47130.02)	−365806.39 (±27076.40)
K = 3, α = 0.01, β = 0.001	−682620.18 (±47179.25)	−365807.88 (±27171.46)
K = 3, α = 0.001, β = 0.1	−686273.30 (±47539.28)	−367994.90 (±27034.03)
K = 3, α = 0.001, β = 0.01	−686666.58 (±47408.23)	−367859.95 (±26874.11)
K = 3, α = 0.001, β = 0.001	−686417.56 (±47156.44)	−367923.44 (±26942.59)
K = 4, α = 0.1, β = 0.1	−677435.77 (±47321.00)	−362748.07 (±26930.41)
K = 4, α = 0.1, β = 0.01	−677681.05 (±46751.86)	−363277.82 (±27292.59)
K = 4, α = 0.1, β = 0.001	−678124.44 (±46816.71)	−363310.62 (±26850.11)
K = 4, α = 0.01, β = 0.1	−678858.29 (±47393.52)	−364019.43 (±27006.19)
K = 4, α = 0.01, β = 0.01	−679569.77 (±47161.13)	−364008.15 (±27215.34)
K = 4, α = 0.01, β = 0.001	−679277.59 (±47150.13)	−364036.90 (±26933.38)
K = 4, α = 0.001, β = 0.1	−683839.21 (±46870.69)	−366149.54 (±26829.01)
K = 4, α = 0.001, β = 0.01	−683417.91 (±46932.56)	−366384.64 (±26828.00)
K = 4, α = 0.001, β = 0.001	−684045.97 (±47494.00)	−366304.03 (±27138.42)
K = 5, α = 0.1, β = 0.1	−674507.71 (±47127.03)	−361290.00 (±26836.91)
K = 5, α = 0.1, β = 0.01	−674681.24 (±47024.50)	−361318.58 (±26818.05)
K = 5, α = 0.1, β = 0.001	−675159.95 (±46797.63)	−361855.70 (±26851.60)
K = 5, α = 0.01, β = 0.1	−676658.40 (±47147.61)	−362468.01 (±27138.36)
K = 5, α = 0.01, β = 0.01	−676662.81 (±47356.85)	−362369.08 (±26737.76)
K = 5, α = 0.01, β = 0.001	−676309.70 (±46958.32)	−362585.89 (±27140.12)
K = 5, α = 0.001, β = 0.1	−681362.66 (±46825.38)	−364723.91 (±27100.00)
K = 5, α = 0.001, β = 0.01	−681469.84 (±47357.16)	−364799.82 (±27106.59)
K = 5, α = 0.001, β = 0.001	−681478.89 (±47564.73)	−364866.80 (±26866.35)

Notice the improvement in log-likelihood achieved by the proposed method when compared to the vanilla LDA model as in ref. ³³.