System and method for efficient machine learning

A system for efficient machine learning with knowledge distillation, comprising:

• a database server executed by one or more storage devices for storing data comprising real-time data and historical data;
• a data analytics system comprising one or more convolutional neural networks;
• and
• a standard dashboard;
• wherein the data analytics system comprises a non-transitory computer-readable medium and instructions that when executed by the non-transitory computer-readable medium cause the non-transitory computer-readable medium to perform operations comprising:
• inputting labeled data, unlabeled data and test data into a computer memory;
• performing an initial self-supervised learning using the labeled data, the unlabeled data, and the test data to output a trained teacher model, wherein the trained teacher model comprises a single base convolutional encoder that is pre-trained using contrastive loss for self-supervised learning;
• supervised finetuning the trained teacher model using the labeled data and the test data to output a finetuned teacher model;
• performing a semi-supervised learning guided by the finetuned teacher model using the labeled data, the unlabeled data, and the test data to output a trained student model;
• acquiring a first data output by the finetuned teacher model and a second data output by the trained student model;
• obtaining a distillation loss value according to a distillation loss function based on a difference between the first data output and the second data output;
• updating parameters of the trained student model based on the distillation loss value until the distillation loss value stops decreasing;
• and,
• outputting a target student model when the knowledge distillation training is completed.