Siddharth Divi

Applied Scientist working on Speech Recognition. Hiker, backpacker, and explorer.

My research spans speech and audio machine learning, federated learning, and time series forecasting. See my full profile on Google Scholar.

Federated Learning

New Metrics to Evaluate the Performance and Fairness of Personalized Federated Learning ICML 2021 Federated Learning Workshop · June 2021

Introduced new metrics to evaluate performance and fairness in personalized federated learning systems beyond average accuracy, enabling more robust comparison of FL algorithms.

Unifying Distillation with Personalization in Federated Learning arXiv · May 2021

Proposed a novel personalization framework for federated learning by unifying knowledge distillation with user-specific optimal teacher models, improving client-level performance in heterogeneous data settings.

Time Series & Forecasting

DeepTrace: A Generic Framework for Time Series Forecasting IWANN 2019 · Gran Canaria, Spain · June 2019

A generic deep learning framework for time series forecasting, presented at the International Work-Conference on Artificial Neural Networks.

Forecasting Food Sales in a Multiplex Using Dynamic Artificial Neural Networks Computer Vision Conference (CVC) 2019 · Las Vegas, USA · April 2019

Applied dynamic neural network architectures to forecast food sales in a multiplex setting.

Species Classification

Convolutional Long Short-Term Memory Neural Networks for Hierarchical Species Prediction CLEF 2018 · Avignon, France · September 2018

Working notes presented at the Conference and Labs of the Evaluation Forum, applying ConvLSTM networks to hierarchical species prediction tasks.