id	author	title	date	pages	extension	mime	words	sentences	flesch	summary	cache	txt
cord-209221-vjfmxsks	Ishiguro, Katsuhiko	Data Transfer Approaches to Improve Seq-to-Seq Retrosynthesis	2020-10-02		.txt	text/plain	7273	458	59	Experimental results show that typical data transfer methods can improve test prediction scores of an off-the-shelf Transformer baseline model. The result shows that every data transfer method can improve the test prediction accuracy of an off-the-shelf Transformer retrosynthesis model. Table 3 : n-best accuracy of retrosynthesis tasks on USPTO-50K, with different data-transfer training methods. n-best accuracy (%) Training Method n = 1 n = 3 n = 5 n = 10 n = 20 n = 50 Single model (No Transfer) 35.3 ± 1.4 52.8 ± 1.4 58.9 ± 1.3 64.5 ± 1.2 68.8 ± 1.2 72.1 ± 1.3 Joint Training 38.4 ± 0.9 60.7 ± 0.5 67.8 ± 0.4 75.2 ± 0.3 80.4 ± 0.4 84.9 ± 0.3 Self-Training 41.2 ± 0.3 60.2 ± 0.4 66.2 ± 0.2 71.9 ± 0.3 75.5 ± 0.5 78.2 ± 0.5 Pre-training + Fine-Tune 52.2 ± 0.4 73.1 ± 0.4 78.8 ± 0.4 83.7 ± 0.3 86.3 ± 0.3 88.2 ± 0.3	./cache/cord-209221-vjfmxsks.txt	./txt/cord-209221-vjfmxsks.txt