This is the Bachelor's thesis for my studies of Data Science at Vilnius University
Learning Transferable Time-Series Models From Natural Language Supervision
This thesis researches relationship between cryptocurrency time series data and Twitter’s tweets about them. By converting time series into images, the original CLIP model is trained to understand new relationships. Then a modified CLIP model is proposed that accepts time series directly without the need for any conversion.
