Computer Vision News - May 2018

order to do the next important step along the path towards making a new drug, for example. Interpretation of deep learning algorithms and models I think is one of the most pressing technical issues. What students should work on is a different question from what does the field need. What the field needs is people willing to collect training sets that accomplish biologically useful tasks. I don’t recommend that necessarily to any particular student or postdoc as a way to spend their time, but it’s one of those thankless tasks that is really critical to progress in the field and yet is a lot of sometimes tedious work. It requires a deep understanding of the biological goals and so on, so that’s a hard thing to ask for, but it’s something we need. The major competitive advantage that anyone has in the field of deep learning has much more to do with the quality and quantity of the data they have available to them, as opposed to any cleverness in their algorithms. Is there anything else you would like to say to our community? I would really encourage your readers - students and postdocs, for example - to look around at what open-source projects they can contribute to in the field. It’s a great way to learn from others, to contribute to society and the computational community at the same time, and to have an impact. If you’re interested in biomedical research, then go out and find some open-source software packages that are heavily used by biologists and biomedical researchers and poke in and lend a hand. One of the interesting things about working on software whose target audience is biologists is that we very rarely have contributions to the software. If you work on a software library whose consumer is other computer scientists, of course they’re going to contribute back - as they find problems they’ll fix them and so on. If you work on software for biologists, who are largely not coders, it’s pretty challenging and lonely. I would encourage your audience to contribute to this kind of software if they want to have an impact. Anne Carpenter ISBI DAILY Friday 27 “… go out and find some open- source software packages that are heavily used by biologists and biomedical researchers and poke in and lend a hand.” BEST OF