Posts Tagged ‘machine learning’
tokyo first impressions
so far all i’ve done is stumbled out of the airport after a 12 hour flight, boarded a train, checked into a hostel, and went to a fast food unagi joint. tomorrow, the famed tokyo fish market and freshest sushi in the solar system. after that, i’m heading to kyoto on the (expensive!) bullet train. the day after that i head to a zen monastery and will stay there sans-internet for 7 days. i emerge on the 11th.
there are more bike riders here than there are in seattle, percentagewise, but i haven’t seen a single road bike. cruisers are standard.
a good portion of the businessmen here have very slim suits. this makes me happy.
the train system is tangled and hard to understand.
the light is somehow different. it is nearer to halo-glow than american light.
i spent most of the airplane ride napping and listening to robert pirsig’s “zen and the art of motorcycle maintenance” on book-on-tape. i have definitely read it before, but i have this weird power to completely forget books, so it was like encountering the ideas within the book for the first time. i like it, even though something about the tone puts me off. the listening sessions did give me some fun ideas to think about. here’s one that concerns machine learning from my notebook:
“problem definition and hypothesis generation are primary acts of human intelligence. modern machine learning does not concern itself with these acts.”
pirsig gave me the idea of thinking about machine learning’s utility in the scientific method. after all, science is our preferred way to learn, which is precisely what machine learning is about. here’s a naive model of the scientific method from wikipedia:
“1. Define the question
2. Gather information and resources (observe)
3. Form hypothesis
4. Perform experiment and collect data
5. Analyze data
6. Interpret data and draw conclusions that serve as a starting point for new hypothesis
7. Publish results
8. Retest (frequently done by other scientists)”
from what i know, machine learning has at least touched on data collection (active learning), data analysis (pretty much the definition of ML), and data interpretation (model selection does this in a limited way). question definition, directed information gathering, and hypothesis & conclusion formation are all outside of its scope. something to keep thinking about.
