woke up and did research again. the talk last night made me realize i never took time to carefuly analyze the data i have, i.e. where it's from, how it was collected, why it was collected that way, whether the data is of good quality, and how to properly structure the data. this is most of the work. model choice, loss function, eval metrics, all come from how good your data is and how you model it. everything follows the data.

spoke to professor for 2 hours considering every aspect of the data and the loss function choice. i'm glad i have a mentor that can support and i can discuss with on the same level. although i've never felt more challenged and frustrated, i've also never grown that much.

this is a reminder to myself to spend more time with the data. if you jump into modeling too quickly, or get too excited about the model architecture, mistakes and issues will come popping up. make sure you understand the ins and outs of your data first.