The Doctor
Senior Member
- Dec 18, 2010
- 1,110
- 474
lolz
How to get into Machine Learning?
- Thread starter DestinyInfinity
- Start date
oseart
Registered Member
- Oct 30, 2015
- 66
- 27
I would LOVE to see a more sophisticated game be played! That would be the coolest thing ever!
ZuzooVn
Junior Member
- Jun 8, 2009
- 104
- 58
Add your aff link on the Amazon links
Lol. I am not going to making any money here. I just want to share my knowledge with everyone.
jazzc
Elite Member
- Jan 27, 2009
- 2,964
- 14,717
One does not exclude the other, does it?
DestinyInfinity
Junior Member
- Oct 3, 2016
- 184
- 63
The Super Mario Video is just awesome. @oseart posted it already, as plain text link. It shows the complex algorithm very good. And com'on, who does NOT like Super Mario?
Would love if I had this 30 years ago, when Super Mario started. Oh my god. What fun.
Update to my OP:
I have looked a little into the topic. As told we have written a social media automation tool. I want to integrate machine learning.
The very first example I want to explain with Facebook. The account has to learn to get as much friends as possible, without getting disabled. As I could see the Reinforcement Learning, or better the Q Learning Algorithm is what we need.
When I am correct, I have to set the agent, environment, states, actions and rewards/score. So far, so clear with the few python examples I watched.
Unfortunately our tool is coded in java, not in python yet.
I spent already some hours but could not find a simple java tutorial for this and a java library with a simple example how to us it.
With python there are awesome tutorials out there, nothing for java?
Would love if I had this 30 years ago, when Super Mario started. Oh my god. What fun.
Update to my OP:
I have looked a little into the topic. As told we have written a social media automation tool. I want to integrate machine learning.
The very first example I want to explain with Facebook. The account has to learn to get as much friends as possible, without getting disabled. As I could see the Reinforcement Learning, or better the Q Learning Algorithm is what we need.
When I am correct, I have to set the agent, environment, states, actions and rewards/score. So far, so clear with the few python examples I watched.
Unfortunately our tool is coded in java, not in python yet.
I spent already some hours but could not find a simple java tutorial for this and a java library with a simple example how to us it.
With python there are awesome tutorials out there, nothing for java?
Last edited:
The Doctor
Senior Member
- Dec 18, 2010
- 1,110
- 474
Here is the easiest Tensorflow tutorial video I could find. If anyone can find an easier one, I'd like to see it
Java is not recommended for ML. The most popular library in Java is DL4J and that lib is only recommended for special projects which must use Java. There's nothing totally wrong with using DL4J but Python is easier and Python libs are on the bleeding edge of development and performance.
Edit: I was wrong about DL4J. There is also Tensorflow for Java: https://www.tensorflow.org/install/install_java
That's what you want to use if you must use Java.
Edit: I was wrong about DL4J. There is also Tensorflow for Java: https://www.tensorflow.org/install/install_java
That's what you want to use if you must use Java.
techn1k
Junior Member
- Mar 4, 2013
- 123
- 35
coldboot
Registered Member
- Nov 30, 2014
- 90
- 28
Try looking at this project. It is not really ML but you can learn something from the work.
Code:
https://github.com/maxbbraun/trump2cashoseart
Registered Member
- Oct 30, 2015
- 66
- 27
Sentdex is probably one of my favorite YouTube channels for learning new things. He has entire series dedicated to one project. I love his GTA V autonomous driver, and his algorithmic bitcoin trader.
DestinyInfinity
Junior Member
- Oct 3, 2016
- 184
- 63
@The Doctor
Thanks. I am already on it with Python. Yesterday I tested some Atari Games with the OpenAI Gym https://gym.openai.com
As I could see there is also a Browser extension, but seems it is only implemented with some basics. Will try it the next days.
OpenAI is also with deep learning, so the code only sees all the pixels and uses reinforcement learning, which is really what I am looking for. Feed the bots just with some small inputs (his goal, his keys he can use and his eyes/window he sees in pixels), and the code does the rest without coding a new intelligence in it.
Very nice stuff. Will look into that.
Thanks. I am already on it with Python. Yesterday I tested some Atari Games with the OpenAI Gym https://gym.openai.com
As I could see there is also a Browser extension, but seems it is only implemented with some basics. Will try it the next days.
OpenAI is also with deep learning, so the code only sees all the pixels and uses reinforcement learning, which is really what I am looking for. Feed the bots just with some small inputs (his goal, his keys he can use and his eyes/window he sees in pixels), and the code does the rest without coding a new intelligence in it.
Very nice stuff. Will look into that.
BusinessSmack
Power Member
- May 2, 2012
- 531
- 163
Great thread. Usually just searching properly in google, youtube or here you find enough answers but I guess this is a new enough topic.
TwoOneThree
Regular Member
- Feb 2, 2017
- 432
- 68
A bit off-topic, but any reason(s) you're usingnot Selenium over bs4?
Currently learning about web scraping and practicing it.
Currently learning about web scraping and practicing it.
thanks for sentdex recommend. is there any other as he have not gone into depth of what i was searching for .
Hello,
I am a full time data scientist so I can help you out with this. Firstly, which languages do you know now? Do you know R or Python?
Also what is your end all goal? Machine learning is pretty broad.
DestinyInfinity
Junior Member
- Oct 3, 2016
- 184
- 63
The coding language doesn't matter. If you have a understanding, you easy learn every language. For the moment I am deep in Java, but also doing some little coding in Python. Our internet bots are written in Java with Selenium.
The end goal depends of course what is possible. I want to tell an example with the openAI like the atari games (also the Super Maio is like this):
Give the bot the environment in Facebook (the browser). Set the goal (fitness level) to maximize the friend counter. Give him the keys/buttons he can press. And then let the openAI try to maximize the friends counter. If the Account gots disabled: The evolution is dead. The bot gets a new Account for a new evolution of the AI.
What I could find is, the openAI has already some basic features for browser machine learning. But did only find some very small disuccion about it and I was not able to run it. I could only run the atari games with the openAI (Gym), not the browser related stuff. Very hard to find anything about it
https://pbs.twimg.com/media/C0G01irXAAAI17H.jpg
https://universe.openai.com/envs#world_of_bits
https://blog.openai.com/universe/ look for "Mini World of Bits" and "Real-world browser tasks" in the middle of the page.
oseart
Registered Member
- Oct 30, 2015
- 66
- 27
There are other people that go into machine learning on YouTube, however I feel as if they don't make it as approachable as he does. It would be helpful if you could explain what you are looking for more in depth.
Oseart
lancis
Elite Member
- Jul 31, 2010
- 1,682
- 2,443
This was the standard during my days in the university:
http://eu.wiley.com/WileyCDA/WileyTitle/productCd-0471056693.html
The book provides a very good mathematical basis of ML.
However, ML is pretty dynamic field so the rest you can search on the need to know basis in scientific publications (such as on arxiv.org).
Learning from YouTube is well... you can consider it as a nice introduction I guess.
http://eu.wiley.com/WileyCDA/WileyTitle/productCd-0471056693.html
The book provides a very good mathematical basis of ML.
However, ML is pretty dynamic field so the rest you can search on the need to know basis in scientific publications (such as on arxiv.org).
Learning from YouTube is well... you can consider it as a nice introduction I guess.
