A constrained recursion algorithm for batch normalization of tree-sturctured LSTM
Tree-structured LSTM is promising way to consider long-distance interaction over hierarchies. However, there have been few research efforts on the hyperparameter tuning of the construction and traversal of tree-structured LSTM. To name a few, hyperparamters such as the interval of state initialization, the number of batches for normalization have been left unexplored specifically in applying batch normalization for reducing training cost and parallelization. In this paper, we propose a novel recursive algorithm for traversing batch normalized tree-structured LSTM. In proposal method, we impose the constraint on the recursion algorithm for the depth-first search of binary tree representation of LSTM for which batch normalization is applied. With our constrained recursion, we can control the hyperparameter in the traversal of several tree-structured LSTMs which is generated in the process of batch normalization. The tree traversal is divided into two steps. At first stage, the width-first search over models is applied for discover the start point of the latest tree-structured LSTM block. Then, the depth-first search is run to traverse tree-structured LSTM. Proposed method enables us to explore the optimized selection of hyperparameters of recursive neural network implementation by changing the constraints of our recursion algorithm. In experiment, we measure and plot the validation loss and computing time with changing the length of internal of state initialization of tree-structured LSTM. It has been turned out that proposal method is effective for hyperparameter tuning such as the number of batches and length of interval of state initialization of tree-structured LSTM.
Code (0)
등록된 구현이 없습니다.
Methods 이 논문이 사용한 방법론
Similar Papers 제목 키워드 기반
Automatically Batching Control-Intensive Programs for Modern Accelerators
We present a general approach to batching arbitrary computations for accelerators such as GPUs. We show orders-of-magnitude speedups using our method on the No U-Turn Sampler (NUTS), a workhorse algorithm in Bayesian sta…
Extended Batch Normalization
Batch normalization (BN) has become a standard technique for training the modern deep networks. However, its effectiveness diminishes when the batch size becomes smaller, since the batch statistics estimation becomes ina…
Image ClassificationAccelerating Monte-Carlo Tree Search with Optimized Posterior Policies
We introduce a recursive AlphaZero-style Monte--Carlo tree search algorithm, "RMCTS". The advantage of RMCTS over AlphaZero's MCTS-UCB is speed. In RMCTS, the search tree is explored in a breadth-first manner, so that ne…
Recursion in Recursion: Two-Level Nested Recursion for Length Generalization with Scalability
Binary Balanced Tree Recursive Neural Networks (BBT-RvNNs) enforce sequence composition according to a preset balanced binary tree structure. Thus, their non-linear recursion depth (which is the tree depth) is just $\log…
Batch Normalization Preconditioning for Neural Network Training
Batch normalization (BN) is a popular and ubiquitous method in deep learning that has been shown to decrease training time and improve generalization performance of neural networks. Despite its success, BN is not theoret…