Resources for Machine learning

[ecoli]

Hi everyone. I'm in a lab right and trying to learn linear algebra, statistics modeling, programming all at the same time and it's overwhelming me. My project is systems biology modeling (regulation network inference) and so it involves learning and clustering. Most of the reviews and textbooks that I'm trying to read are too advanced or too slow.

 

Does anybody know of any texts or resources where I learn this stuff gently, but in a way where I wouldn't have to essentially do a whole applied math degree just to get up to scratch?

 

I started watching the Kahn Academy lectures on linear algebra, but it's fairly slow. Strang's course on MIT OCW and his applied LA textbook is a little better, but again it's more general than what I really need.

 

Any recommendations or should I just suck it up and spend a lot more time learning this stuff to get a comprehensive education.

[Xittenn]

There are plenty of others with Math degrees that are putting out platforms so that individuals like yourself can do work without such knowledge. I think the answer to this question greatly depends on what it is specifically you hope to accomplish.

 

I would recommend that if there is a specific application that you wish to develop that you might simply learn how to use the Matlab Machine Learning Tools or possibly learn Haskell and try out some of the developed libraries; maybe even learn to wrap Haskell into Matlab.

 

I have no degree but I am looking to go to UBC for a degree in Bio. I have decided that for me the best way is to maintain a thorough study in mathematics and computer sciences and then apply my knowledge in mimetics. Rigor for me will mean that I will have the capability of designing a system from root where all the technologies will be specialized to my specification.

 

It is really pendent on what you hope to accomplish and who is paying the bill. If you are in a lab setting then you are most likely restricted by a finite amount of time and financial resources. What does your employer hope that you accomplish? It can't hurt to play super genius if you have the time and desire!

 

There is no specific literature, as you have defined, that I have seen or can recommend on this topic and it is something I have been into for a very long time!

[ecoli]

I'm in an academic setting, and will likely be taking courses in the future, but a self-study background would be helpful for more immediacy. Eventually I'd like to be able to design/program my own systems, but for now I'm using tools developed by others in the lab.

[CharonY]

Simplified there are three aspects to it. First, the mathematical principles and generalized modeling principles, second the models that have been developed in the frame work of one systems theory or other and third is software development to do the numerical calculations.

 

I am doing mostly the second part (i.e. looking and adapting existing models to my research questions, usually by using or adapting existing software/code) to supplement certain aspects of my research. While the first part is certainly best to build a solid foundation I am too much experimental to delve into it too much. If you want to develop novel approaches it would certainly help (but it would be rather ambitious).

 

I can provide lit for applied approaches and models if you could provide information what you are interested in. I am less suited to evaluate basic principles.

[Xittenn]

An Introduction to Neural Networks

By James A. Anderson

 

I read this book in High School, it says a lot about what it is that you might wish to achieve inside of a model. It's not a book specifically on Machine Learning, I had picked it up to learn about neural networks and to gain insight into fuzzy systems. I'm sure there are books that better describe the process of learning from a programmatic point of view, like doing lazy walks on graphs as matrices and so on. This does sort of sketch out the foundation of what the representation is trying to achieve! Maybe?

Edited March 18, 2011 by Xittenn

[khaled]

An Introduction to Neural Networks

By James A. Anderson

 

I read this book in High School, it says a lot about what it is that you might wish to achieve inside of a model. It's not a book specifically on Machine Learning, I had picked it up to learn about neural networks and to gain insight into fuzzy systems. I'm sure there are books that better describe the process of learning from a programmatic point of view, like doing lazy walks on graphs as matrices and so on. This does sort of sketch out the foundation of what the representation is trying to achieve! Maybe?

 

Machine Learning is a field, you need to learn basics on the following:

 

- Fuzzy Logic (good knowledge in Logic too)

 

- Statistics & Probability Mathematics (Mathematical basics in Calculus)

 

- Modeling & Simulation

 

- Neural Networks

 

.. good luck