Here is a compact representation of binary decision trees. We explicitly formulate the dependence of prediction on binary tests for decision trees and construct a procedure to guide the input sample from the root to its exit node. And we provides a connection between decision trees and error-correcting output codes. Then we borrow the ideas from attention mechanism to approximate and extend this formulation via continuous functions.

翻译：我们提出了一种二叉决策树的紧凑表示方法。明确阐述了决策树中预测结果对二元测试的依赖性，并构建了一个引导输入样本从根节点到达出口节点的流程。同时建立了决策树与纠错输出编码之间的关联。进一步借鉴注意力机制的思想，通过连续函数对该表述进行近似与扩展。