Hello! I have a bunch of training data and am playing with machine learning, but have no experience with TensorFlow and thus am looking for someone to show me the ropes. Specifically I am looking for someone to:
1) Answer some general questions and give advice
2) Build a proof of concept atop TensorFlow in C++
3) Deploy on Google ML-Engine or Amazon Lamda (perhaps by porting to a different language)
Regarding stage 1, once you accept the bid, I will give a more comprehensive description of the exact problem domain (eg, what data we have and what results are looking for), and then based on that I am looking for you to answer the following questions:
a) How do you recommend mapping our non-image input problem domain to a TensorFlow model?
b) Which Estimator do you recommend for our application, and why? Please list all of the "standard" estimators and provide a description of why each would or would not be appropriate.
c) Do you recommend using TensorFlow locally, on Google's ML-Engine, or on Amazon Lamda (or something else), and why?
d) What is the most parallel way to train our Estimator? Is it possible to train a single model using multiple input/output pairs in parallel, or is the best practice instead to just train many separate models in parallel (even though each is trained sequentially) to determine which performs best -- ala genetic programming?
As for how stage (1) will work, let's just schedule a time where we chat (via a private Slack room) and you bill me hourly. Don't worry, I'll make pretty efficient use of your time, and will pay you handsomely for it. Once we get through stage (1), we will figure out the best way to approach billing for stages (2-3) -- I'm open to hourly or fixed bid, as you prefer.
Less than 30 hrs/week
1 to 3 months1-3 monthsProject LengthDuration
I am willing to pay higher rates for the most experienced freelancers