English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

This paper develops a general framework for learning interpretable datarepresentation via Long Short-Term Memory (LSTM) recurrent neural networks overhierarchal graph structures. Instead of learning LSTM models over the pre-fixedstructures, we propose to further learn the intermediate interpretablemulti-level graph structures in a progressive and stochastic way from dataduring the LSTM network optimization. We thus call this model thestructure-evolving LSTM. In particular, starting with an initial element-levelgraph representation where each node is a small data element, thestructure-evolving LSTM gradually evolves the multi-level graph representationsby stochastically merging the graph nodes with high compatibilities along thestacked LSTM layers. In each LSTM layer, we estimate the compatibility of twoconnected nodes from their corresponding LSTM gate outputs, which is used togenerate a merging probability. The candidate graph structures are accordinglygenerated where the nodes are grouped into cliques with their mergingprobabilities. We then produce the new graph structure with aMetropolis-Hasting algorithm, which alleviates the risk of getting stuck inlocal optimums by stochastic sampling with an acceptance probability. Once agraph structure is accepted, a higher-level graph is then constructed by takingthe partitioned cliques as its nodes. During the evolving process,representation becomes more abstracted in higher-levels where redundantinformation is filtered out, allowing more efficient propagation of long-rangedata dependencies. We evaluate the effectiveness of structure-evolving LSTM inthe application of semantic object parsing and demonstrate its advantage overstate-of-the-art LSTM models on standard benchmarks.

Interpretable Structure-Evolving LSTM