Logic could be learned from images

Logic reasoning is a significant ability of human intelligence and also an important task in artificial intelligence. The existing logic methods, quite often, need to design some patterns beforehand. This has led interesting question: can be directly learned from given data? problem termed as data concept logic. In this study, learning images, called LiLi task, first proposed. learn reason the relation without presetting any patterns. As preliminary exploration, we six sets (Bitwise And, Bitwise Or, Xor, Addition, Subtraction Multiplication), which each image embedded with n-digit number. It worth noting that model beforehand does not know meaning numbers images between input output image. order tackle work use many typical neural network models produce fruitful results. However, these have poor performances on difficult task. For furthermore addressing novel framework divide conquer by adding label information designed, achieving high testing accuracy.