Systems and methods for training multimodal self-supervised models

Abstract

A system for training multimodal self-supervised models, the system including a computing device configured to receive a multimodal dataset encode each data entry of a plurality of data entries into a single data format to create an encoded multimodal dataset including one or more encoded modalities, identify one or more embedded pairs as a function of the one or more encoded modalities, wherein each embedded pair of the one or more embedded pairs is associated with a separate patient of the plurality of patients, map the embedded pairs within a shared embedding space, pretrain an ECG machine learning model as a function of the mapping by adjusting one or more parameter values of the ECG machine learning model, receive labeled training data and train the pretrained ECG machine learning model as a function of the labeled training data.