Bioinformatics 2012-02-07

Bioinformatics
2012-02-07
Stanford

“Let me remind everybody where we were last time.”

“Variants to the model.”

histogram

“The point here is not just to be mathematically elegant.”

“One very elegant way to do this…”

“The most important question: How does this help you? How can you use this?”

“Is there a biological explanation for this mathematical phenomenon?”

negative binomial

assigning function to nucleotide polymorphism

Wow.

“I don’t even have an F11.”

This is pretty amazing.

“I can draw this a little different on the board.”

“Disadvantage: Increase in complexity.”

“Some of you may actually have read.”

You remember things best when you do them, not when you write them.

“Sorry I have it a little bit differently here.”

“Please forget what I just said.”

“What I want to say is…”

“If I were to be precise…”

“I forgot to write that.”

model the duration based on the emission frequency

“I think the answer to this is yes.”

Don’t be afraid to express confusion.

“OK, so what’s the complexity of this algorithm?”

O(k^2*n*d)

“This is still much more computational demanding than the original algorithm.”

Sometimes being fast works better than being good.

“Precompute cumulative values.”

“Maybe…I’m trying to make an example now.”

“I want to talk about learning.”

“We are not going to talk about the structure of learning in this class.”

Two learning scenarios:
1. Estimation when the “right answer” is known
2. Estimation when the “right answer” is unknown

The second is derived from the first.

practice in your head

“Then we are pretty much done.”

“Given little data, there may be over-fitting.”

“Very heuristically.”

“Here is the idea.”

You have to feel dumb to get smarter.

“We can keep repeating this until convergence.”

“I don’t know if you agree but the whiteboard is always better than the slides.”

“We have some random ones, but we have them.”

“In fact, we can do something more.”

“Do you see what I did? Do you all see what I did?”

No one can do the work for you.

Work with a small team that solves real problems.

“Make sense?”

“OK, now what have we accomplished?”

Everybody likes compliments.

“I’m not going to do this.”

“You are actually asking a question that is very very relevant to the thing I next wanted to say.”

cross validation

“It’s useful to try. I’m not particularly fond of it.”

“In general, worse performance.”

You can’t just surround yourself with data, you have to test it.

“Why do they have those names here? I don’t know.”

“Typically the way you would use this is…”

“We would train on a few of our examples and test on the rest.”