machine learning spider

algorithm and data are both objects

In supervised learning, data.X is input, while data.Y is output

Two important functions:
[data2 ...]=train(algorithm, data, ...)
[data2 ...]=test(algorithm, data, ...)

data2.X is not the same as data.X, but the prediction of data.Y

data2=test(algorithm,data,'class_loss');
only calculate class loss, which is stored in data2.Y
Its equivalent is
data2=test(algorithm,data);
data3=loss(data2,'class_loss');


其他重要 object
algorithm 有很多类，包含 feature normalization，feature selection, classifier, etc. 这些都用 train 函数。
train(normalize,d)
chain 是串联一系列 algorithms, train(chain({algo1 algo2 algo3}),data)
相当于
d2 = train(algo1,data)
d3 = train(algo2,d2)
d4 = train(algo3,d3)

group 类似于 cell

• chain
• param
• group
• loss
• cv
• get_mean
• grid_sel