Table 2.

Comparison of PCT and TgDPF under different input noise N(0, x) in training data. The results are evaluated using the MSE metric, reflecting the models’ predictive accuracy under noisy input conditions.

Noise		Exp 0	Exp 1	Exp 2	Exp 3	Exp 4	AVG
	TgDPF	0.0322	0.0307	0.0300	0.0301	0.0319	0.0310
N(0, 0.1)	PCT (Ours)	0.0264	0.0250	0.0272	0.0264	0.0259	0.0262
	Impro(%)	18.0	18.6	9.3	12.3	18.8	15.4
	TgDPF	0.0353	0.0328	0.0329	0.0323	0.0327	0.0332
N(0, 0.2)	PCT (Ours)	0.0298	0.0277	0.0283	0.0297	0.0287	0.0288
	Impro(%)	15.6	15.5	14.0	8.0	12.3	13.3
	TgDPF	0.0374	0.0356	0.0372	0.0376	0.0385	0.0373
N(0, 0.3)	PCT (Ours)	0.0345	0.0337	0.0348	0.0349	0.0358	0.0347
	Impro(%)	7.7	5.3	6.5	7.2	7.0	6.7
	TgDPF	0.0407	0.0427	0.0405	0.0398	0.0390	0.0405
N(0, 0.4)	PCT (Ours)	0.0396	0.0408	0.0406	0.0392	0.0389	0.0398
	Impro(%)	2.7	4.5	−0.2	1.5	0.3	1.7
	TgDPF	0.0466	0.0464	0.0463	0.0443	0.0439	0.0455
N(0, 0.5)	PCT (Ours)	0.0567	0.0553	0.0599	0.0657	0.0657	0.0607
	Impro(%)	−21.7	−19.2	−29.4	−48.3	−49.7	−33.4