Machine learning books

#1
Hey guys,

im currently writing my masters thesis using predictive tools for toxicity (structure activity relationship as well as all kinds of QSAR and statistical systems) ,so far I have been mostly only using them without any background knowledge, or at least only very little from what i got of the publications.
Now I would love to advance further into Machine learning as I am very interested in doing a PhD towards Machine learning in life sciences ( got an Msc in Toxicology with great interest in Bioinformatics).
I tried to do as much towards programming as I could but did not get much further than basic Java and well slightly advanced stuff in R.
I would love to get a book which gives an Introduction to machine learning and which might not be too hard for non-mathematicians to understand just to get some principal knowledge and to see whether this topic would actally be an option for my PhD.

Hope some of you have suggestions for me!
 

hlsmith

Not a robit
#2
This book is a great introduction and is free:


http://www-bcf.usc.edu/~gareth/ISL/


Though it is very basic, and does not focus on Neural Networks and some other approaches. The authors also have another book that follows this one called Elements of Statistical Learning. I also think Andrew Ng is in the process of releasing soon an overview book.
 

TheEcologist

Global Moderator
#4
Bishop 2006 is the authoritative book on the topic, it is comprehensive and I like it. I'm however more experienced in the subject and this book may be best after you have enjoyed a softer introduction on the topic.

Bishop, C. Pattern Recognition and Machine Learning (2006). A pdf is floating around on the web for those who know how to use that famous search engine.
 
#5
Hey thanks to you as well!
It looks like this books covers the topic more in depth (at least from a first view) and might be the follow up to teh first book ;)
I dont know what your background is but mine is just very low level maths (2 semesters of maths for natural sciences) thus i guess ill need slightly more thinking when reading the mathematical descripton. but ill go for it and try =)

As i guess i will be merely using the systems and choose appropriate methods how much of the statistics du you think i should understand?
 
Last edited:

TheEcologist

Global Moderator
#6
As i guess i will be merely using the systems and choose appropriate methods how much of the statistics du you think i should understand?
As much as you can. Not only does decent knowledge of statistics aid in the analysis it also ensures a better study design to start with. A good comprehensive understanding will also allow you to tweak the methods and adapt the system to your needs increasing your power. Your ability will be limited by your understanding. Yet all this does depend on the time you can invest initially.
 
#7
Bishop 2006 is the authoritative book on the topic, it is comprehensive and I like it. I'm however more experienced in the subject and this book may be best after you have enjoyed a softer introduction on the topic.

Bishop, C. Pattern Recognition and Machine Learning (2006). A pdf is floating around on the web for those who know how to use that famous search engine.
thanks. I'm also in search of book, could help me in math understanding. but its difficult as compare to my level. any other?