06 · Published Research

Smaller ensembles, useful uncertainty

A shared-representation ensemble architecture that improves predictive diversity without repeating an entire network.

Role

Author · machine learning research

Year

2024

Diagram showing shared and divergent branches in an ensemble neural network

In one sentence

The work was published in the International Journal of Artificial Intelligence & Applications.

01 · The problem

Why this mattered

Deep ensembles are valuable for uncertainty estimation, but independently training every model is expensive and redundant.

02 · My work

From idea to system

With Advait Chandorkar, I developed Divergent Ensemble Networks: a common representation with independent branches designed to preserve useful disagreement.

03 · The result

What moved

The work was published in the International Journal of Artificial Intelligence & Applications.

Next projectA robot arm that learns by watching