Debate of Torch Lua vs Tensor Flow Keras for ultimate machine learning
There seems to be a debate of these two frameworks:
- 1. Redis 4 with Neural module and embedable Lua scripts make this combo ultimate fast with no latency unlike TensorFlow as a bloated framework.
- 2. TensorFlow is accused of being slow so you can compare by searching online for it.
- 3. Will Torch in Lua be much faster than TensorFlow in Python? Apparently a big yes from the Reddit postings I have seen. It is just not as mature which means migrating a potential model into a live trading environment might not be so easy from TensorFlow. This is the issue I face.
- What to do?
As I approach a more serious deeper machine learning investigation, I will see where everything is at then which could be a couple of more months. Torch may be more mature with better integration with Redis.
- It might be best to use the potential Redis TensorFlow module combo but that seems premature. Find the link below. I will wait it out to see how this Redis module develops.
Keras with TensorFlow
https://github.com/fchollet/keras/issues/4491
https://deeplearning4j.org/compare-dl4j-torch7-pylearn
Zerobrane studio for Lua:
http://notebook.kulchenko.com/zerobrane/torch-debugging-with-zerobrane-studio
Redis NOSQL options:
https://news.ycombinator.com/item?id=12617081
Potential combo with Redis and TensorFlow https://github.com/lantiga/RedisTF
http://redis-knowledge.fgribreau.com/resource/2e8908b15a864927e78e5424cf67d63c/redis-c-integration-layer
Here are some links on speedy Torch
Lua options
http://lua-users.org/wiki/SimpleStats
NOTE I now post my TRADING ALERTS into my personal FACEBOOK ACCOUNT and TWITTER. Don't worry as I don't post stupid cat videos or what I eat!