Publications

Personalised Federated Learning

The similarity-based sorting and grouping federated learning (SiRGFL) algorithm is proposed to be a unique method used to compute the similarity of model performance across all clients. Client-specific performance data is collected through pre-training of federated learning and the iterative data is displayed in a high-dimensional space and downscaled as well as permutated to obtain a federated learning grouping ranking that balances the model performance with generalizability.

The paper is under review.

Anomaly Detection

Develop an adaptive image detection system in the robotics lab to analyze multimodal data and integrate an intelligent recognition system based on the “anomalib” library to quickly detect unknown anomalies.

The paper is under review.

Bofinformatics

Using convolutional neural networks for denoising and phasing mixed haplotypes from nanopore sequence data in Plasmodium falciparum.

The Paper is under review.