Hi Tanya, Welcome to the Shogun community! (please always cc the mailing list - I know it can be intimidating to have so many people see your emails, but sure enough there will be students/Shogun mentors out there who have similar questions/know the answers to your question/want to know about this)
Thank you for you interest and the many ideas! Your background also sounds very suited :) Yes, the applications project is still “available” - we only make decisions about student applications after the official deadline (27th March). However, to make sure your application has a chance for success, please note the steps in our “How to get involved”: - contribute Code (it is important for us to see Code contributions from you *before* the application deadline. This helps us to judge both your dedication and also expertise so we know, e.g. who might be best to mentor you. WE DO NOT ACCEPT APPLICATIONS WITHOUT PRIOR CODE CONTRIBUTION) - don’t just contact individual mentors but get active in the community (answer questions, report bugs, contribute) - write a well thought-through project proposal. While it is great that you have so many ideas, what you need for a GSoC project is a detailed plan (including, importantly, a timeline) that you can follow through with. I am happy to give you several rounds of feedback on your application before the deadline but your ideas below are too vague for me to see where a project could be headed (and in particular: where Shogun comes into play in these ideas). So decide on one, spell it out and then let’s have a discussion about it! Looking forward to your contributions and your project proposal! Best, Lea > On 22 Mar 2018, at 13:06, merlettaia <[email protected]> wrote: > > Hi Lea, > > I want to ask if data application project > (https://github.com/shogun-toolbox/shogun/wiki/GSoC_2018_project_data_application > > <https://github.com/shogun-toolbox/shogun/wiki/GSoC_2018_project_data_application>) > is still available and any of my ideas fits. > I understand that probably I'm too late, but the main reason for this is > because this year I realized that I'm eligible to GSoC, got excited, couldn't > choose and forgot about it :/ > That's what I can do (or transform this to something similar): > 1. "Popular-science bot" > Basically this is not a one idea, this is a set of two similar ideas which > solves 2 problems. > (a) It transformed from the problem my and my friends usually deal with - a > lot of science papers. Originally I wanted to implement something similar to > this: > https://github.com/titipata/science_concierge > <https://github.com/titipata/science_concierge> > and to implement some form of filtering to propose new papers to read based > on current research topic, methods I use and and labs I admire. > But this was unclear and I postponed implementing it. > (b) Last fall I found this article: https://arxiv.org/abs/1711.00043 > <https://arxiv.org/abs/1711.00043> > and thought that it would be great to implement something interesting - for > example, to use aligned language embeddings try to translate from scientific > language to common-language (popularize articles). I've tried to implement > this at Junction, hackathon held in Helsinki, Finland, but couldn't do this > in 2 days, since I couldn't prepare dataset in time and my AWS account was > granted access to servers with GPU only after the whole event finished :) > I had some concerns regarding data licensing, but for now I know for sure > that there are several popular science news sites at least in Russian with > CC-BY license (for Junction I wanted to collect data from English resources) > with links to related articles at PubMed (in English), and I understand how > to collect these data. > > 2. VAENGA-SMILES: > With my coworkers we had an idea to implement variational autoencoder (VAE) > for small molecules generation and properties prediction guided by genetic > algorithms, but will probably postpone it till the end of times, since my MSc > student, who is chemist, and who was expected to write genetic > algorithms-related part, . > > 3. I'm a PhD student, I study structural bioinformatics and for my research > I've chosen to try to apply geometric deep learning for protein's active site > generation for given (protein) target. This seems crazy, since I plan to > model the surface, and there is also structure (and some aminoacid sequence) > behind, and I still didn't figure out how I would reestablish it. > So, as a PhD student, I'm interested in developing a VAE-based algorithm, in > ideal case - to learn to use geometric deep learning. > I didn't started this yet, since in my university there are studies that > takes some time, and last term I worked on quite different project by request > of my PhD advisor with our collaborators (on metagenomics, and mostly I run > tools on their data - as a result, we have a talk accepted to Recomb-seq). > The GSoC project would be another reason not to give up and to work on my > project. > > Last fall I found this article: > https://www.autodeskresearch.com/publications/exploring_generative_3d_shapes > <https://www.autodeskresearch.com/publications/exploring_generative_3d_shapes> > - it describes car meshes generation with VAE. > I think it would be useful for my project to reproduce this or implement > similar pipeline and run it on other data (to make this project idea not too > nerdish). > > If you like any of this, I could provide more precise description or details, > and will do starting contribution. If not, please tell me. > > -- > Kind Regards, > Tanya Malygina. > > P.S. - this is my CV for you to understand what I can do: > https://docs.google.com/document/d/1m00WeLcf6VEM6aQroXiHTKZAUNj8P4x9TZRWhOhHNeE/edit?usp=sharing > > <https://docs.google.com/document/d/1m00WeLcf6VEM6aQroXiHTKZAUNj8P4x9TZRWhOhHNeE/edit?usp=sharing> > In short, I'm a former Outreachy program member at Debian, currently I work > at biomedical startup (+enrolled to PhD). Take a look at "Pet projects" > section - there are several projects I'm for some reasons proud of.
