使用字典替换 NumPy 数组中的值会给出模棱两可的结果，这是为什么呢？

Question

So, I have an array with some words an I'm trying to perform one-hot encoding.

Let's say the input is AI DSA DSA AI ML ML AI DS DS AI C AI ML ML C

This is my code:

def apply_one_hot_encoding(X):
    dic = {}
    k = sorted(list(set(X)))
    for i in range(len(k)):
        arr = ['0' for i in range(len(k))]
        arr[i] = '1'
        dic[k[i]] = ''.join(arr)
    
    for i in range(len(X)):
        t = dic[X[i]]
        X[i] = t
         
    return X

if __name__ == "__main__":
    X = np.array(list(input().split()))
    
    one_hot_encoded_array = apply_one_hot_encoding(X)
    for i in one_hot_encoded_array:
        print(*i)

Now, I would expect the output to be like:

1 0 0 0 0 
0 0 0 1 0 
0 0 1 0 0 

But what I'm getting is:

1 0 0
0 0 1
1 0 0

If I append the t values to another list and return that, it is giving the right results.

Why is the assigned value being trimmed to just 3 characters in case of direct substitution?

Answer 1

The problem is caused due to the dtype (datatype) of the Numpy array.

When you check the datatype of the numy array in the above program using print(X.dtype) , it shows the data type as <U3 which can hold only three characters for each element in the numpy array X .

Since input array contains five categories, the dtype of the array can be changed into <U5 by X = np.array(list(input().split()), dtype='<U5') which can hold upto five characters for each element in the numpy array X .

The corrected code is,

def apply_one_hot_encoding(X):
    dic = {}
    k = sorted(list(set(X)))
    for i in range(len(k)):
        arr = ['0' for i in range(len(k))]
        arr[i] = '1'
        dic[k[i]] = ''.join(arr)
    
    for i in range(len(X)):
        t = dic[X[i]]
        X[i] = t
         
    return X

if __name__ == "__main__":
    X = np.array(list(input().split()),dtype = '<U5')
    
    one_hot_encoded_array = apply_one_hot_encoding(X)
    for i in one_hot_encoded_array:
        print(*i)

The above method is not needed when you store the values in the separate numpy array since the numpy changes the datatype automatically according the size of strings,