Reed Naidoo's Projects
A sound data multilabel classification problem for 153 species of birds.
Retraining a SegFormer with a branch that generates query and key embeddings from intermediate hierarchical feature map logits. This is done so that a contrastive learning task can be performed alongside the segmentation model.
Markov Chain Monte Carlo used for posterior probability distribution prediction. Under the Bayesian paradigm, two algorithmic updating approaches are explored - one in which we know the parameters of the underlying distribution of our data, and one in which we don't know the parameters of the underlying distribution of our data.
An introduction to tensorflow in R. Notebook includes a code, that for any mutilabel dataset, with predictions, it computes a Macro-AUC as well as plots for those curves.