…. As Maslow said: “If the only tool you have is a hammer, you tend to see every problem as a nail.” At Riskified we deal with structured, domain-driven, feature-engineered data which is best dealt with various forms of boosting trees. Having someone whose entire CV points back to DL is an issue.
Nice to hear this, I started losing interest in DL as of lately because to me it's a lot like turning knobs and hoping for the best. The fact that I know backprop doesn't mean much as it's so far removed from any end model anyway. Plus I avoid it every time I can because it's not great for the environment (energy consumption...)
I'm going over all of classical ML again. I like to be able to have more understanding of what's going on in each model and turning knobs knowing what will happen. And with classical ML, I think there are no shortcuts. A lot of DL-only people learn Tensorflow and Keras and call themselves machine learning engineers. I'd rather have a deep knowledge of classical ML only. If nothing else, it makes me stand out.
Thanks for the tips! I just got a new job but always good to review my CV!