At the end of this chapter
Participating in this program was a very rich experience for which I am very grateful. It was a pleasure knowing my mentors and Tanya, my project partner. Throughout the whole summer Dr. Gall helped us a lot in creating the models and instructed about what we should do. He knows a lot and he taught us more about Artificial Intelligence and Neural Networks. Dr. Pitoňák helped us in writing the report and making the video at the end of the project and gave us some good feedback.
I am very grateful to all of them for taking some time of their summer, which usually people use for vacations and other things, to be our mentors and for giving us such attention. Tanya and I were always helping each other and exchanging our progress and our results. I am glad she took the challenge of working on this project and that we could work together.
Although I knew some of the theory, I started the summer without having any practical experience with neural networks. I knew much about the concepts, but I had never used them in an actual project. Yet in this project, we used Neural Networks to create all our models, therefore I could immerse myself into the tools. After learning more about Keras and tensorflow I am much more confident and excited to work on future projects that use NNs to create models. In case you are interested in exploring what we did in detail, please check our final report. The video bellow is a presentation that we prepared to explain some of the work we developed.
Now I know better some of the practical challenges of NN and AI. In theory, these techniques seem very useful. Yet, when I was working with them I realized that they also offer some challenges. Finding a proper set of parameters is one of them. Sometimes there are so many parameters to tune that it is very difficult to explore all possible combinations. And once it is very difficult to tune all these parameters, it is also very unlikely to be certain that our model is the best and the tuning is optimal.
We also experienced how long the codes can take to run and how paralyzation can help. I had never faced these things before since the models I dealt with were much simpler and used much lesser data in the training process. Moreover, as we had access to different computers, it was interesting to see how the hardware can influence the time needed to run the codes. We got access to very impressive machines which was something I never had experienced before. Although I believe I couldn’t reach some of the initial goals, I am very satisfied with my overall experience with this project. Eight weeks can be short to learn about a new topic and also develop a project. Yet, I am impressed with everything I learned during this short period.
Even though we are finishing this project, I hope this will be only the end of a chapter for me. I believe that the techniques for molecule design that we learned will be the future of most chemical product engineering. And these techniques are just some of all the others there are available. As a chemical engineering student, I am very excited to deepen even further in the field. Certainly, the things that I learned during this summer will help me in my career, and it already influenced my master thesis.
For the students that are thinking about applying for the future of HPC, I totally recommend it. If anyone wants to learn more about what we did or more about my experience with Summer of HPC, or just wants to talk about something related to molecule engineering, feel free to contact me.
I want to thank everyone from PRACE for the awesome program that they offer during the summer. I also want to thank everyone that supported me during this journey, such as Isabela, Giulia, Daniela, Roberto, Juliana, and other friends and relatives. This is just the beginning of my journey through molecule engineering using computational tools.