Ricerc@Sapienza

There is a variety of available approaches to learn graph node embeddings. One of their common underlying task is the gener- ation of (biased) random walks that are then fed into representation learning techniques. Some techniques generate biased random walks by using structural information. Other approaches, also rely on some form of semantic information. While the former are purely structural, thus not fully considering knowledge available in semantically rich networks, the latter require complex inputs (e.g., metapaths) or only leverage node types that may not be available.

Graph embedding techniques allow to learn high-quality feature vectors from graph structures and are useful in a variety of tasks, from node classification to clustering. Existing approaches have only focused on learning feature vectors for the nodes and predicates in a knowledge graph. To the best of our knowledge, none of them has tackled the problem of directly learning triple embeddings.

While word embeddings are now a de facto standard representation of words in most NLP tasks, recently the attention has been shifting towards vector representations which capture the different meanings, i.e., senses, of words. In this paper we explore the capabilities of a bidirectional LSTM model to learn representations of word senses from semantically annotated corpora. We show that the utilization of an architecture that is aware of word order, like an LSTM, enables us to create better representations.

We study maximal identifiability, a measure recently introduced in Boolean Network Tomography to characterize networks' capability to localize failure nodes in end-to-end path measurements. Under standard assumptions on topologies and on monitors placement, we prove tight upper and lower bounds on the maximal identifiability of failure nodes for specific classes of network topologies, such as trees, bounded-degree graphs, d-dimensional grids, in both directed and undirected cases.