Posts by Collection

portfolio

Semi-Supervised Learning with few labels

Published:

Semi Self-Supervised Learning: improving the performance of self-supervised learning models, especially in scenarios where only a small amount of labeled data is available

GliZNet: Generalized Zero-Shot Text Classification

Published:

A novel zero-shot multi-label text classification architecture that embeds labels directly in the input sequence, achieving efficient classification through supervised contrastive learning and label repulsion.

publications

Detecting Misinformation and its Sources on Social Media

Published in -, 2022

This work proposes an efficient solution for detecting and filtering misinformation on social networks, specifically targeting misinformation spreaders on Twitter during the COVID-19 crisis, using a Bidirectional GRU model that achieved a 95.3% F1-score on a COVID-19 misinformation dataset, surpassing state-of-the-art results.

Recommended citation: Alex Kameni, 2022
Download Paper | Download Slides

DATA INCREMENTAL LEARNING IN DEEP ARCHITECTURES

Published in -, 2022

This study presents a framework for continual self-supervised learning of visual representations that prevents forgetting by combining distillation and proofreading techniques, improving the quality of learned representations even when data is fed sequentially.

Recommended citation: Alex Kameni, 2022
Download Paper | Download Slides