Naredni sastanak Seminara biće održan onlajn u sredu, 5. marta 2025. godine, sa početkom u 19 časova.

Predavač: João Pedro Carvalho Leal Mendes Moreira, Faculdade de Engenharia da Universidade do Porto (FEUP)

Naslov predavanja: AN ENSEMBLE OF AUTONOMOUS AUTO-ENCODERS AS A CLASSIFIER FOR HUMAN ACTIVITY RECOGNITION

Apstrakt: Human Activity Recognition is a machine learning task that uses sensing technology to classify human activities and infer human behavior. While traditional machine learning approaches use hand-crafted features to train their models, recent advancements in neural networks allow for automatic feature extraction. Auto-encoders are a type of neural network that can learn complex data representations and are commonly used for anomaly detection. In this work, we propose a novel multi-class algorithm comprising an ensemble of auto-encoders, with each auto-encoder associated with a unique class. We tested the proposed approach in human activity recognition data. Experimental results show that auto-encoder ensembles can be efficient, robust, and competitive with traditional classification approaches. Moreover, this modular classifier structure allows for more flexible models. For example, the number of classes can be extended by including new auto-encoders without retraining the whole model.

Napomena: 
Registraciona forma za učešće i link za aktivno praćenje predavanja za registrovane korisnike (nakon logovanja):
https://miteam.mi.sanu.ac.rs/asset/CW5nJWDSEZDj7p32p

Neregistrovani korisnici mogu da prate predavanja na ovom linku (bez mogućnosti aktivnog učešća):
https://miteam.mi.sanu.ac.rs/call/hR9vL94nD6QE8qQZj/xET9GcPMyR08nqH8lnS3SE7N5Vf00H7Lp9EBhsv6Lti