d892fb853c4c296539034d2dbaf304c0d06a357d,qiskit_acqua/ising/graphpartition.py,,sample_most_likely,#Any#Any#,180

Before Change


    if isinstance(state_vector, dict) or isinstance(state_vector, OrderedDict):
        temp_vec = np.zeros(2**n)
        total = 0
        for i in range(2**n):
            state = np.binary_repr(i, n)
            count = state_vector.get(state, 0)
            temp_vec[i] = count
            total += count
        state_vector = temp_vec / float(total)

    k = np.argmax(state_vector)
    x = np.zeros(n)

After Change


    
    if isinstance(state_vector, dict) or isinstance(state_vector, OrderedDict):
        // get the binary string with the largest count
        binary_string = sorted(state_vector.items(), key=lambda kv: kv[1])[-1][0]
        x = np.asarray([int(y) for y in list(binary_string)])
        return x
    else:
        n = int(np.log2(state_vector.shape[0]))
        k = np.argmax(np.abs(state_vector))
Italian Trulli
In pattern: SUPERPATTERN

Frequency: 4

Non-data size: 6

Instances


Project Name: Qiskit/qiskit-aqua
Commit Name: d892fb853c4c296539034d2dbaf304c0d06a357d
Time: 2018-07-06
Author: 34400304+liupibm@users.noreply.github.com
File Name: qiskit_acqua/ising/graphpartition.py
Class Name:
Method Name: sample_most_likely


Project Name: scikit-learn-contrib/categorical-encoding
Commit Name: 695b48f181fd35751da2f5406373003866d4838d
Time: 2019-06-13
Author: slliu96@163.com
File Name: category_encoders/hashing.py
Class Name: HashingEncoder
Method Name: transform


Project Name: scikit-learn-contrib/categorical-encoding
Commit Name: be362c73ceabd21bc2dd72218c7a97c950ef786c
Time: 2019-06-13
Author: slliu96@163.com
File Name: category_encoders/hashing.py
Class Name: HashingEncoder
Method Name: transform


Project Name: Qiskit/qiskit-aqua
Commit Name: a194557ba754f9b14d473ff9e39a2bc2449e58c1
Time: 2018-07-06
Author: chenrich@us.ibm.com
File Name: qiskit_acqua/ising/maxcut.py
Class Name:
Method Name: sample_most_likely