GSoC stuff

If you’re someone who’s here because of GSoC related things, here’s some stuff I wish I knew before starting out:

  • First, remember that you will have to go out of your comfort zone when building stuff.

  • Your mentor is your best friend through all the odds.

  • The better you know the objective of the org, the better your proposal will be. Dig around and find out the approximate thought process that went behind the formation of the project if possible.

  • Make sure you’re ready to learn new things. In my case, I had to learn a lot of things on the fly.

  • The experience I gained from my personal projects in the past helped me the most.

  • On writing proposals:

    • Proposals should not be about what you know, it should be about what you plan to build with what you know.

    • Pictures/visualizations > words

    • If possible, show them that you’ve done something similar to the project before.

    • Less jargon, more reasoning.

    • Show them how exactly is your approach better than the average approach. If your approach has some downfalls, mention them as well. This would mean that you know what you’re doing.

    • Original projects that actually work > courses. This specifically applies to deep learning related GSoC projects like the one I did.

Miscellaneous

It’s been the most productive 3 months I’ve ever had in my life. When I say productivity, I do not mean pushing code. Productivity is when I’m learning and applying new things, not when I’m just doing something I’ve done before. If I’m not making mistakes, I’m not learning new things.

As much as I don’t want so sound like some nerd, I’m not productive unless I’m backpropagating on my own mistakes. The best way to do that is to first calculate the loss by reflecting on the mistake and analysing where I went wrong.

new_action = old_action - learning_rate*gradient

where:

The right amount of patience is like the adaptive learning rate that everyone needs. Too fast and you overshoot, too slow and you’ll never reach the minima.

Thank you for reading this far.

“It is better to fail in originality than to succeed in imitation.”

– Herman Melville