Which is the Recommended variants of neural network for tree like data structures? Consider I have a training set where each data sample is a tree and each node has(including leaf node) its own feature vector. For example, single data sample will look this

Now I have to classify leaf nodes to either of two categories(category_1 or category_2). Can some one please recommend feasible neural network variant for this kind of data.
 A: You might be in interested in, in contrast to a recurrent neural network, a recursive neural network, which

[applies] the same set of weights recursively over a structure, to produce a structured prediction over variable-size input structures, or a scalar prediction on it, by traversing a given structure in topological order ... introduced to learn distributed representations of structure.

You may also be interested in parsing in general, in which tree structures are inferred for sentences, for example.  Each tree structure follows certain rules specified by some grammar.  
One paper that might be of interest is aclweb.org/anthology/P/P13/P13-1045.pdf
I have not used them myself, but it appears that there are implementations in pytorch (https://devblogs.nvidia.com/parallelforall/recursive-neural-networks-pytorch/) and tensorflow  (https://www.kdnuggets.com/2016/06/recursive-neural-networks-tensorflow.html).
A: There is this paper on Neural Decision Trees. Where the decision tree is the network structure and each node is a perceptron. I also seem to call another paper by researchers in China that made headlines within the last 6 months but what I read about that one was that is was mostly hype. 
