Comparative study of machine learning and deep learning methods on ASD classification

The autism dataset is studied to identify the differences between autistic and healthy groups. For this, the resting-state Functional Magnetic Resonance Imaging (rs-fMRI) data of the two groups are analyzed, and networks of connections between brain regions were created. Several classification frameworks are developed to distinguish the connectivity patterns between the groups. The best models for statistical inference and precision were compared, and the tradeoff between precision and model interpretability was analyzed. Finally, the classification accuracy measures were reported to justify the performance of our framework. Our best model can classify autistic and healthy patients on the multisite ABIDE I data with 71% accuracy.

Sparse Popularity Adjusted Stochastic Block Model

The objective of the present paper is to study the Popularity Adjusted Block Model (PABM) in the sparse setting. Unlike in other block models, the flexibility of PABM allows to set some of the connection probabilities to zero while maintaining the rest of the probabilities non-negligible, leading to the Sparse Popularity Adjusted Block Model (SPABM). The latter reduces the size of parameter set and leads to improved precision of estimation and clustering. The theory is complemented by the simulation study and real data examples.