python策略覆盖for循环中的全局变量

Question

EDIT2:

Thanks for your help, problem solved, went with an intermediate approach:

Will accept the answer when it becomes eligible to be accepted!

---

EDIT:

I was asked for simpler variables, okay, I will try :)

Assume I want to have data in separate variables a , b , c , and d , how would I assign data to them in a for loop?

An approach like this:

a, b, c, d = [], [], [], []

for var,data in zip([a,b,c,d], some_data_array):
    var = #data that comes from some operation on data array

wouldn't work, the variables a, b, c, d would still be [] , since they are in the global scope.

---

I have a training dataset train_set (numpy array) that I want to reduce into smaller training datasets (90%, 80%, 70%, etc.). I don't want to copy & paste the procedure

train_90perc = train_set[np.random.choice(train_set.shape[0],\
        train_set.shape[0] * 0.9, replace=False),:]

and thought I could do it with a for loop. But how? My initial approach (below) has the problem that it wouldn't modify the variables in the global scope. So when I would print eg, train_80perc after the for loop, it would still be [] . PS: The first 2 lines are to initialize the variables, is there also a way I can get around that (since it looks kind of ugly).

train_100perc, train_90perc, train_80perc, train_70perc, train_60perc, train_50perc,\
train_40perc, train_30perc, train_20perc, train_10perc = [], [], [], [], [], [], [], [], [], []

for train,size in zip([train_100perc, train_90perc, train_80perc,\
                       train_70perc, train_60perc, train_50perc,\
                       train_40perc, train_30perc, train_20perc, train_10perc],\
                       [p for p in reversed([i/10 for i in range(1,11)])]
                     ):
    #global train
    train = train_set[np.random.choice(train_set.shape[0],\
    train_set.shape[0] * size, replace=False),:]

Thanks for your help!

Answer 1

UPDATE

Don't assign some data to var . because it's just reference.

a, b, c, d = [], [], [], []

for var,data in zip([a,b,c,d], some_data_array):
    # call var.extend or var.append
    # var = some_value # don't do this. because it's just reference.
    var.extend(some_list)

---

Do you have to generate train_100perc ... ?

My suggestion is

subset = {}
for percent in np.arange(10, 100, 10):
    np.random.shuffle(train_set) # shuffle in-place
    subset[percent] = train_set[:len(train_set)*percent*0.01].copy() # selection of training data

np.arange(10, 100, 10) can vary.