New Technique Identifies Novel Class of Cancer's Drivers

New Technique Identifies Novel Class of Cancer's Drivers

It was only a matter of time that modern DNA sequencing methods would be able to uncover some "use" of the enormous amount of non-coding DNA in our genome. We still have so much to learn! But, this is a great step.

How people learn - The Week

I'm posting this article for my own reminder about the importance of teaching in a way that maximizes student learning.

I've made a remarkable number of changes over the past year with my teaching tools and techniques. Yet, I still find myself occasionally drifting back to old habits -- giving a standard, boring lecture where I expect my students to sit there, pay attention and take notes for 52 minutes of class. Fortunately, this is certainly becoming the exception and not the norm.

Regardless, research is continuing to emerge about how ineffective this traditional approach is on the majority of students. However, I don't believe it is ineffective for everyone. After all, the vast majority of my undergraduate years were all completed by taking courses where I had to sit through lectures where the professor delivered the material for the entire class with talking and an overhead as their primary tools. I did it. It worked for me, yet I can't help but wonder how much more enjoyable some of my undergraduate years would have been if I had some active learning....

Big data, big jobs?

See the article Big data, big jobs? in a recent issue of ComputerWorld.

I teach an undergraduate course in data mining. It is definitely one of my favorite courses to teach.  I've been telling students over the past couple of years that this area is a huge opportunity in the job market, and will continue to blossom for years to come.

Yet, even as this article suggests, data mining is not for everyone.

Here are my observations: Unfortunately, I am not seeing enough students ready or willing to dive into some of the statistical math required to really understand how to get the most interesting information out of the data.  Being a good programmer is not enough for getting into data mining. Knowing a good programming language that is popular for data mining tasks (such as R) will be better.  However, knowing how to analyze, visualize, and extract relevant information from large data sets is important. To do this, you need to know what models to apply when, and why, and what to expect out of them.  You need to know how to evaluate model performance, and select appropriate parameters to improve it.  You need to understand causes of poor performance (e.g. noisy data, lack of preprocessing, etc.) I see too many students blindly meandering around the data mining landscape with programs such as SAS, SPSS, or even Weka, not really understanding why a particular model is behaving the way it is. (BTW, Weka is an absolutely wonderful piece of software for exploring data mining!)

The best foundation you can give yourself for a career in data mining is to give yourself a solid foundation in statistics and probability. Then, take a course in data mining! I find that failure to fully embrace these important topics makes it difficult for a student to understand the strengths and weaknesses of the wide range of statistical models and algorithms for inference and induction in data mining tasks. More importantly -- you will likely miss many hidden gems buried in the data, and this is what your potential employer is after.

How Software Updates Are Destroying the Stock Market - Businessweek

This is a sobering reminder of how far we have yet to go in this modern era of software development of engineering. We continue to strive to release some great development tools, and yet it is practically impossible to set up tests to capture every possible event that might raise exceptions....

High school student wins Google Science Fair with an artificial neural net to predict breast tumor malignancy

You can view her presentation and information related to the project here. I am thoroughly impressed. Congratulations Brittany on a job well done! You are going to go extremely far!

Now, I have two questions:

1. How can I recruit her to work with me on my research?  The reality is that Brittany is most likely MIT or CMU bound, though she will be able to choose whatever school she really wants, and will most likely get a full scholarship on top of the awards she has already earned. Awesome.
   
2. How do I get in touch with her parents to see what they did with this amazing girl as she was growing up? My son is almost 4 yrs old now. And, right now, he is all about Thomas the Tank Engine! But, the young mind is so malleable, impressionable... they are little sponges! Do I start giving him some Java and Python books now? (Yes, I'm kidding... I'll wait until he is 5 yrs old. :-) Do I get him started with some neat robotics project? Teach him how to code in R? Yet, I want him to enjoy his life as a child. This is more important to me. It goes by so fast! There must be a balance, right? I suspect that the voice of encouragement and support from the parental units are probably the most critical part of any child's success. But, is there something more? Perhaps feeding them more brain food? Keeping them away from the TV? I wonder what steps parents need to take to maximize our chances of raising more Brittanys for this world? What did they do to inspire such motivation and creativity in her mind at such a young age?
   
   To me, it sounds like this is an opportunity for data mining to answer these questions... if I only had a large repository of data...

Again, congratulations Brittany on a job well done! Awesome job, girl. Your parents must be extremely proud of you.