• Publications
  • Influence
Modeling documents with Generative Adversarial Networks
A model that is based on the recently proposed Energy-Based GAN, but instead uses a Denoising Autoencoder as the discriminator network is proposed. Expand
Evaluating the Supervised and Zero-shot Performance of Multi-lingual Translation Models
An in-depth evaluation of the translation performance of different models, highlighting the trade-offs between methods of sharing decoder parameters, finds that models which have task-specificDecoder parameters outperform models where decoder parameter are fully shared across all tasks. Expand
Task Selection Policies for Multitask Learning
This work provides an empirical evaluation of the performance of some common task selection policies in a synthetic bandit-style setting and on the GLUE benchmark for natural language understanding, and suggests a method based on counterfactual estimation that leads to improved model performance in experimental settings. Expand
Real-time detection of musical onsets with linear prediction and sinusoidal modeling
Comprehensive results for both the detection accuracy and the computational performance of all of the described techniques are provided, evaluated using Modal, the new open source library for musical onset detection, which comes with a free database of samples with hand-labelled note onsets. Expand
Real-time segmentation of the temporal evolution of musical sounds
Since the studies of Helmholtz, it has been known that the temporal evolution of musical sounds plays an important role in our perception of timbre. The accurate temporal segmentation of musicalExpand
360{\deg} Stance Detection
360{\deg} Stance Detection, a tool that aggregates news with multiple perspectives on a topic that presents them on a spectrum ranging from support to opposition, enabling the user to base their opinion on multiple pieces of diverse evidence. Expand
N ov 2 01 8 Off-the-Shelf Unsupervised NMT
We frame unsupervised machine translation (MT) in the context of multi-task learning (MTL), combining insights from both directions and draw connections to policy-based reinforcement learning. WeExpand
SIMPL: A Python Library for Sinusoidal Modelling
Simpl, a new open source library for sinusoidal modelling written in Python, is presented as a resource for researchers in spectral signal processing who might like to access existing methods and techniques. Expand
Metamorph: Real-Time High-Level Sound Transformations Based On A Sinusoids Plus Noise Plus Transients Model
Spectral models provide ways to manipulate musical audio signals that can be both powerful and intuitive, but high-level control is often required in order to provide flexible real-time control overExpand