Mysql dump and restore tables

To dump tables t1, t2, ... from mysql dababase db:
mysqldump -u USERNAME -p db t1 t2 ... > tables.sql

To restore the tables:
mysql -u USERNAME -p db < tables.sql

Python loop dictionary

d = {}

Python 2: for key, val in d.iteritems():

Python 3: for key, val in d.items():

python - No module named MySQLdb?

Simply using commend: pip install MySQL-python

Something to the notice for "foreach $elemrnt (@array)"

In each foreach iteration, if you change one element, the array will also be changed!!!

Perl sort function

Sort in alphabetical order, using 'cmp'; sort in numerical order, using '<=>'.

Sort an array:

    my @new =  sort {$a cmp $b} @old;

Sort a hash (sort by key, return sorted keys):

    my @new =  sort {$a cmp $b} keys %old;

Sort a hash (sort by value, return sorted keys):

    my @new =  sort {$old{$a} cmp $old{$b}} keys %old;

Reasons for over-fitting and how to fix it

Usually, over-fitting is caused by the ability to learn complex hypothesis, and aggregated by the noise in the training data.

To fix over-fitting, we can: (1) add more training data, (2) simply the training model, (3) prune the data, (4) use gold annotation to test the model got, stop training when error goes up.

Some terminology is NLP

Induction: from example, generate rules.

Deduction: from rules, get instance.

Bias: inability of model function to approximate data. Usually arise from not sufficiently complex hypothesis space.

Variance: sensitivity to change in training data. Usually due to having the ability to learn complex hypothesis.