Organize and bin you say? I wonder if such a task might interest the Tensorflow Tuesdays group (@mrcity, @mrhavens, @Josh_Melnick, @otto1)…
Yeah, electronics classification has been on my mind for a while, but it might be more useful for a novice trying to learn resistors or their way around a board rather than trying to help us bin parts. I think by the time you get enough electronics experts together, and considering all the different bags & tubes these parts usually exist in now, it’d go pretty fast.
2 Likes
I actually have a brilliant idea for Machine learning at the space.
Since we already have a hadoop, elk stack, and tensorflow clusters plus the weather station nonstop collecting data. Why not combined it all and do some weather prediction models?
1 Like
Or we could do some Bayesian classification models on some spam/ham sets?
1 Like
The spam/ham would certainly be easier… however there used to be this really old game on Apple 2 or early Macintoshes called “Five Star Forecast”. I thought it’d be fun to bring this back as some kind of MMO game and see if there are individuals among us who can issue a better forecast than your average TV meteorologist with any regularity.
My vote is start small unless there are already open-source weather prediction models that we could just plug in. I’m just the Tensorflow deployment guy, not the math whiz 
2 Likes
open-source weather prediction models
you where asking? https://arnesund.com/2017/02/26/using-facebook-prophet-forecasting-library-to-predict-the-weather/
If I’m reading it right then its just taking the data sets from the weather station, plugging it into scipy+numpy+pandas and letting tensorflow do the calculations for the forecastpredictions or at lease cleaning up the probability of a prediction based on learning from the delta of predicted weather “flavor (ie properties of wind, temp, and other measurements)” vs actual values of collected time serial data.
Mind you, I’m thinking of this from a statistical mindset not AI developer. Neuronets, hash comparisons, cryptography, and Classifications when applied to natural language, event correlations, and code analysis is about as far I’ve ever touch in the realm of AI/ML data sciences or even “big data” kind of stuff.
Though it would be great to build out some predictive models at the space and maybe later apply it to some AI first design apps.
So, tell me more about this “Five Star Forecast”? What do you envision for it?