Portfolio

A Multivariate Hawkes Process for Detecting Individuals with Depressive Disorder

We propose a personalized framework for classifying individuals with and without depression symptoms. We simulated the interplay between the distributional semantics and the temporal stochasticity of online activities with a mutually exciting multidimensional Hawkes Process.

Machine Learning → Temporal Point Process; Mental Health.

Understanding Diversity-based Pruning of Neural Networks – Statistical Mechanical Analysis

Despite the multitude of empirical advances, there is no theoretical understanding of the effectiveness of different pruning methods. We address this issue by setting up the problem in the statistical mechanics formulation of a teacher-student framework and deriving generalization error (GE) bounds of specific pruning methods.

Theory for Deep Learning → Statistical Mechanical Analysis.

Decoding Word Sense with Graphical Embeddings

Implemented three deep graph encoders over WordNet for word sense embeddings bounded by hypernyms-hyponyms and meronyms-holonyms: a child-sum graph bi-LSTM, a matrix decomposition method with graph kernels, and a GCN.

Word Sense Disambiguation.