Faster and more accurate machine learning techniques with less data

Abstract

With the latest development of deep learning techniques and ability to process the data in acceptable timeframe, the need to consider the aspects of environmentally-friendly machine learning techniques has arose. In addition, the latest development of IoT technologies led to the trend where the data are collected and actively used in various machine learning techniques. The lecture will explorer how to reduce the computational requirements of machine learning techniques during training, how to identify the completeness of dataset and ensure that only “useful” data have been used to enhance the training models? How can we design environmentally-friendly machine learning that requires minimum CO2 and respectively computational resources?