Computational Intelligence and Neuroscience

Research Article

TAFM: A Recommendation Algorithm Based on Text-Attention Factorization Mechanism

Results of ablation experiment.


Model variants	Explanation

TAFM_ae	TAFM_ae refers to the removal of the convolutional autoencoder from the TAFM model.
TAFM_a	TAFM_a refers to the removal of the outermost multiheaded attention mechanism based on the TAFM model.
TAFM_ae_a	TAFM_ae_a refers to the TAFM model with both the outermost multiheaded attention mechanism and the convolutional autoencoder removed.
TAFM_l	TAFM_l refers to the TAFM model with the LSTM removed.
TAFM_b	TAFM_b refers to the removal of BiLSTM_Attention from the TAFM model.
TAFM_c	TAFM_c refers to the TAFM model with the textCNN removed.
TAFM_l_b	TAFM_l_b refers to the TAFM model with both LSTM and BiLSTM_Attention removed.
TAFM_l_c	TAFM_l_c refers to the base of TAFM model with both LSTM and TextCNN removed.
TAFM_b_c	TAFM_b_c refers to the removal of both BiLSTM_Attention and TextCNN on top of the TAFM model.
TAFM_b_l_c	TAFM_b_l_c refers to the TAFM model with BiLSTM_Attention, LSTM and TextCNN removed at the same time; in other words, this variant removes the entire text module.