5minbreak

I posit that complex NN models can achieve the same level of interpretability as logistic regression. In part, because some interpretability methods use logistic regression as a white box model to explain black boxes.

In other words: If you are comfortable OK'ing a logistic regression model (because you looked at the coefficients and they made sense), you should be comfortable OK'ing a complex NN model (because the evaluation and interpretability modeling makes sense).

Nitpicking, but significant: Most models don't output decisions, they output predictions. Decision scientists then build a policy on top of the model. Key issue here is that the policy makers don't trust the predictions. But I posit they have no reason to trust the predictions of a logistic regression model any more than the predictions of a complex black box. Provided, of course, you deliver interpretability UX, confidence estimates, and strong statistical guarantees and tests. Which is possible for even the blackest of boxes.

If automatic justification is impossible for computers/black boxes, I believe it is impossible for humans too (as per Church-Turing). But let's say it is impossible. Do you think Google would use a white box model to optimize Adsense, because they can't interpret powerful deep learning solutions (like risk management for BlackRock: a very critical part of their business)?

I'd say Google came pretty close with https://distill.pub/2018/building-blocks/ (they are not the only players in the interpretability field, and plenty of methods are becoming available, in large part driven by academia not business: interpretability and fairness are not too important for the bottom line).

Both a fault of the employees that worked on this, and the manager.

Your deliverable should be an interpretable model. You can (and probably should) make neural network models interpretable. If upper management does not trust your performance evaluation enough to bet on it, either the evaluation was weak (and no model should be deployed, however simple and interpretable) or upper management doesn't know enough about modern ML to have to make these decisions.

I have sympathy for the manager in charge for making a decision on a complex model (while all they ever knew was simple survival models and basic statistical models). But you got to move with the times. Your competitors will use the most powerful models available (and some may go under due to improper risk management). Your employees don't want to build logistic regression models until eternity.

k-nearest neighbors classification is one of the first non-linear supervised learning algorithms. Its predictions are derived from the data sample distances.

It is basically a glorified fuzzy lookup table, but then again, so can one view deep learning (fuzzy hierarchical localized lookup).

Pure memorization can even outperform logistic regression, especially with big data sets, so there is some recent debate as to what degree models memorize and to what degree they generalize.

Yeah, I cared about this issue. I think it is an interesting direction that tech is going in, and to see some pushback against the identity politics/social justice movement.

Perhaps comparable to changing the name Black Hat so to not offend people of color. Reasonable or madness?

Agree on 2. If only someone in the US administration really hated what he calls "Fake News" by established media and "Made-up anonymous sources".

"Won't anybody think of the poor economy! These unfounded stories are doing direct damage. Disclose your sources to the government or lose your Real News License."

https://en.wikipedia.org/wiki/%F0%9F%8D%A0 progress (HN does not like emoji's it strips them: cool URL's don't change)

Likely influenced in part by survivor bias. I don't have a PhD either. You can say I have a more practical than theoretical understanding of dropout.

The way to get past HR gatekeepers is to bypass HR. If you ask an employee: "Hey, I'd really like to work at your company for a while now. Can you share some resources/tips that would help me prepare for function x?" and you don't get useful feedback, you don't want to work there. If you send a data science lead a notebook where you solve a problem that is relevant to their business and you don't get a job interview, you don't want to work there (or need to brush up on your presentation and analytics skills).

Do the fast.ai courses. I promise you there are more profitable jobs available in ML than in indie game dev. Become good at using ML tools and data science (Python) and/or data engineering and data infra (Scala, AWS, Docker).

Game development always scared me as one of those studies that you hope your kids don't get interested in. Another one is manual drawing, especially when it is Anime (what are the chances that someone from a small European country is going to excel at such a job?).

> Abandoning the game completely doesn’t seem like a sane option, after all the time I put in.

Addictive games use the above cognitive bias to make you keep coming back, trying to level up by grinding boring quests. "You are not done yet, but the end is in sight!".

As a good game dev, in 3 years, you can: Learn TensorFlow. Publish an AI paper. Beat state of the art on a few datasets.