Hello there! It has been a while, and a lot has happened in terms of project development. My team and I are working on putting all the pieces together that we individually worked on into one large automated bitcoin trading bot! My portion of that, as previously discussed, is that I built our reinforcement model (a DQN to be more specific). But what did I do to come up with this model?
Well, first and foremost was a bunch of reading. I had very little experience developing or implementing reinforcement models prior to this, so learning more about the different model archetypes and then the more specific features of a DQN (including what a DQN even is) took up a large chunk of the time for this project.
Secondly, when it came to the actual coding of the implementation page, a lot of that came from referring to the Keras documentation page. There was not too many relevant examples to pull ideas from in common reference places like StackOverflow, so I had to go to the source.
Additionally, I tried out using ChatGPT to bounce some ideas around and determine what the best fit was. It didn’t really provide me much assistance when it came to the actual implementation or debugging purposes, but it was very helpful in bringing in new concepts or ideas to explore further. So I used it more as a sort of thesaurus, I would feed it some idea that I wanted to explore and it would provide me with some related information that I would investigate further. It actually ended up being a huge help when it came down to deciding on the final model design.
However, most of the implementation trouble is on the horizon. Once we have to start getting the whole thing to run after putting all the pieces together I can already see the runtime errors popping up. So let’s see how we sail to the finish line!
– Joe
Leave a Reply