Parameters | Models | Descriptions |
---|---|---|
fixed η and λ (λP = λP) model | M1 | SGD. A fundamental SGD-based LFA model that converges very fast with the resilience for the HDI data analysis |
M2 | Moment SGD. For a better update of the current one in the process of the SGD-based LFA model, it incorporates the previous updates into the current one in the way of momentum | |
M3 | Adam. This adaptive moment estimation SGD method further considers exponentially decaying of square average and past stochastic gradients average | |
η-λ-(λP = λP) adaptive model | M4 | ALF. An adaptive SGD-based LFA model that adopts standard PSO to adjust the learning rate in the model |
M5 | DELF. Differential evolution-based LF analysis model employs original differential evolution to conduct an SGD-based LFA model adaptively | |
η-λ-(λP ≠ λP) adaptive model | M6 | The proposed ADMA. Compared with DELF, the proposed model ADMA improves the original DE algorithm for a more precise prediction in an SGD-based LFA model with limited time |