wlamond

A digital picture frame. The one I have gets a custom email address that family can use to send me pictures, or they can use an app to add pics. It's been great having pics of siblings, nieces and nephews, parents, etc all popping up on my frame. It does what FB used to do for me before the feed was filled with random things that aren't from my friends and family.

Thanks for the article! What tool do you use to create the figures? I like the sketch style.

For anyone else, what tools do you recommend for generating pretty for system architectures, workflows, etc.?

He's saying there are 1000 possible PINs (000-999) if you don't have the unique digit constraint.

I bet we could. The length of each utterance, the number of exchanges between agents, and the entropy of the symbols used in the utterances could give you some measure of efficiency.

It'd be interesting if the agents developed their own language during the reinforcement learning stage that is unintelligible to humans but allows them to quickly navigate the negotiation. They use the model trained in a supervised way during the reinforcement learning stage to avoid this, but I'm curious to see what the agent learns when paired against another reinforcement learning agent.

Edit: Indeed, the paper says that not using the fixed agent trained on human negotiation leads to unintelligible language from the agents.

The data should have been split into train, validation, and test sets before augmentation was applied. As it is now, there are likely examples in the validation and test sets that are highly similar to examples in the training set. The goal of validation should be to measure performance on previously unseen patients, not previously unseen scans of patients you've already trained on.

Given the marketing toward HPC tasks, shouldn't hyper-threading be disabled?

"And then (this is hard for coders) trusting the systems to do the work."

Like you say, it can be easy to think that something works when it really doesn't. I hope that the above quote isn't meant to be interpreted as "believe the results are correct." Evaluation is paramount when working on these systems to avoid making such mistakes. I assume Google is including evaluation in their machine learning training, but it would have been nice to see that pointed out in the article for folks who may have an interest in machine learning but don't know what's important to focus on.

I was in my first data structures class when this paper was published. We had just covered red black trees, so this made for an interesting deviation from the standard curriculum. Considering the maturity of the data structures covered in the standard undergrad CS curriculum, I wonder how often this type of situation comes up. That is, how often are papers on new data structures published that are this close to something a typical undergrad student studies?

Sort of. There are out-of-core learning algorithms (e.g. SGD) that work on data that doesn't entirely fit in main memory. Please see: http://scikit-learn.org/stable/auto_examples/applications/pl...

I don't work at Google, but I think they do that to gather click through data for links and positions on the results page. Simply linking to the destination would prevent them from learning valuable information about the quality of their search engine and what users think is valuable.