EqNet-L: A new representation learning method for mathematical expression
Jiaxin Liu · 2022 International Conference on Big Data, Information and Computer Network (BDICN) · 2022
In the field of machine learning, symbolic reasoning has not made substantial progress in the past decades compared to deep learning. Researchers believe that combining deep learning may help, and the key is that the neural network needs to learn the features in the mathematical expressions. EQNET was explicitly created for the representation learning of mathematical expressions. Although it has made significant progress in learning representations of mathematical expressions, we believe that there is still much work that can be done. On this basis, we propose EQNET-L. Its structure is mainly based on EQNET, and We applied Dropout in network, the module STACKED-SUBEXPAE is proposed to make sure the network could learn more semantic information. Compared with the previous models, EQNET-L achieved better results in training and testing with the same hyperparameters as EQNET.