Nested LSTMs

• Main Authors: Moniz, Joel Ruben Antony, Krueger, David
• Format: Journal Article
• Language: English
• Published: 31.01.2018
• Subjects: Computer Science - Computation and Language; Computer Science - Learning
• DOI: 10.48550/arxiv.1801.10308

Proceedings of the Ninth Asian Conference on Machine Learning, PMLR 77:530-544, 2017 We propose Nested LSTMs (NLSTM), a novel RNN architecture with multiple levels of memory. Nested LSTMs add depth to LSTMs via nesting as opposed to stacking. The value of a memory cell in an NLSTM is computed by an LSTM cell, which has its own inner memory cell. Specifically, instead of computing the value of the (outer) memory cell as $c^{outer}_t = f_t \odot c_{t-1} + i_t \odot g_t$, NLSTM memory cells use the concatenation $(f_t \odot c_{t-1}, i_t \odot g_t)$ as input to an inner LSTM (or NLSTM) memory cell, and set $c^{outer}_t$ = $h^{inner}_t$. Nested LSTMs outperform both stacked and single-layer LSTMs with similar numbers of parameters in our experiments on various character-level language modeling tasks, and the inner memories of an LSTM learn longer term dependencies compared with the higher-level units of a stacked LSTM.