3
$\begingroup$

I am new to python programming and Qiskit. I would like to ask can we use sklearn for adding external .csv data file before Qiskit libraries, or without sklearn, we can do or is there any other Qiskit library to add any external .csv file, to run the python code of quantum support vector machine?

How I can use .csv data file with Qiskit libraries to run quantum SVM ?

I am following the code written below, where I should put my own .csv data file

import numpy as np
from dataset import breast_cancer
from sklearn.datasets.samples_generator import make_blobs
from qiskit.aqua.utils import split_dataset_to_data_and_labels 
from sklearn import svm
from utils import svm_utils 
from matplotlib import pyplot as plt
#matplotlib inline
#load_ext autoreload
#autoreload 2   % Breast Cancer dataset

n = 2 # number of principal components kept
training_dataset_size = 20
testing_dataset_size = 10

sample_Total, training_input, test_input, class_labels = breast_cancer(training_dataset_size, testing_dataset_size, n)

data_train, _ = split_dataset_to_data_and_labels(training_input)
data_test, _ = split_dataset_to_data_and_labels(test_input)  # %%Breast Cancer dataset

##%% Linear Support vector machine
# We use the function of scikit learn to generate linearly separable blobs
centers = [(2.5,0),(0,2.5)]
x, y = make_blobs(n_samples=100, centers=centers, n_features=2,random_state=0,cluster_std=0.5)


fig,ax=plt.subplots(1,2,figsize=(10,5))


ax[0].scatter(data_train[0][:,0],data_train[0][:,1],c=data_train[1])
ax[0].set_title('Breast Cancer dataset');

ax[1].scatter(x[:,0],x[:,1],c=y)
ax[1].set_title('Blobs linearly separable');

##%% Hands-on session on support vector machine

plt.scatter(data_train[0][:,0],data_train[0][:,1],c=data_train[1])
plt.title('Breast Cancer dataset');

model= svm.LinearSVC()
model.fit(data_train[0], data_train[1])
accuracy_train = model.score(data_train[0], data_train[1])
accuracy_test = model.score(data_test[0], data_test[1])


X0, X1 = data_train[0][:, 0], data_train[0][:, 1]
xx, yy = svm_utils.make_meshgrid(X0, X1)
Z = model.predict(np.c_[xx.ravel(), yy.ravel()])
Z = Z.reshape(xx.shape)

fig,ax=plt.subplots(1,2,figsize=(15,5))
ax[0].contourf(xx, yy, Z, cmap=plt.cm.coolwarm)
ax[0].scatter(data_train[0][:,0], data_train[0][:,1], c=data_train[1])
ax[0].set_title('Accuracy on the training set: '+str(accuracy_train));

ax[1].contourf(xx, yy, Z, cmap=plt.cm.coolwarm)
ax[1].scatter(data_test[0][:,0], data_test[0][:,1], c=data_test[1])
ax[1].set_title('Accuracy on the test set: '+str(accuracy_test));

##%% We now implement a SVM with gaussian kernel
clf = svm.SVC(gamma = 'scale')
clf.fit(data_train[0], data_train[1]);
accuracy_train = clf.score(data_train[0], data_train[1])
accuracy_test = clf.score(data_test[0], data_test[1])


X0, X1 = data_train[0][:, 0], data_train[0][:, 1]
xx, yy = svm_utils.make_meshgrid(X0, X1)
Z = clf.predict(np.c_[xx.ravel(), yy.ravel()])
Z = Z.reshape(xx.shape)

fig,ax=plt.subplots(1,2,figsize=(15,5))
ax[0].contourf(xx, yy, Z, cmap=plt.cm.coolwarm)
ax[0].scatter(data_train[0][:,0], data_train[0][:,1], c=data_train[1])
ax[0].set_title('Accuracy on the training set: '+str(accuracy_train));

ax[1].contourf(xx, yy, Z, cmap=plt.cm.coolwarm)
ax[1].scatter(data_test[0][:,0], data_test[0][:,1], c=data_test[1])
ax[1].set_title('Accuracy on the test set: '+str(accuracy_test));

##%% First steps in Qiskit
import qiskit as qk

# Creating Qubits
q = qk.QuantumRegister(2)
# Creating Classical Bits
c = qk.ClassicalRegister(2)

##%% Define and print empty circuit
circuit = qk.QuantumCircuit(q, c)
print(circuit)

##%%% Add gates to the qubits
# Initialize empty circuit
circuit = qk.QuantumCircuit(q, c)
# Hadamard Gate on the first Qubit
circuit.h(q[0])
# CNOT Gate on the first and second Qubits
circuit.cx(q[0], q[1])
# Measuring the Qubits
circuit.measure(q, c)
print (circuit)

##%% Run the circuit on the quantum simulator
# Using Qiskit Aer's Qasm Simulator: Define where do you want to run the simulation.
simulator = qk.BasicAer.get_backend('qasm_simulator')

# Simulating the circuit using the simulator to get the result
job = qk.execute(circuit, simulator, shots=100)
result = job.result()

# Getting the aggregated binary outcomes of the circuit.
counts = result.get_counts(circuit)
print (counts)

from qiskit.aqua.components.feature_maps import SecondOrderExpansion

feature_map = SecondOrderExpansion(feature_dimension=2, 
                                   depth=1)

##%%Print the feature map circuit
x = np.array([0.6, 0.3])
print(feature_map.construct_circuit(x))

##%%% QSVM Algorithm
from qiskit.aqua.algorithms import QSVM

qsvm = QSVM(feature_map, training_input, test_input)

##%% Run QSVM
from qiskit.aqua import run_algorithm, QuantumInstance
from qiskit import BasicAer

backend = BasicAer.get_backend('qasm_simulator')
quantum_instance = QuantumInstance(backend, shots=1024, seed_simulator=10598, seed_transpiler=10598)

result = qsvm.run(quantum_instance)

##%%% Analyze output
plt.scatter(training_input['Benign'][:,0], training_input['Benign'][:,1])
plt.scatter(training_input['Malignant'][:,0], training_input['Malignant'][:,1])
plt.show()
length_data = len(training_input['Benign']) + len(training_input['Malignant'])
print("size training set: {}".format(length_data))
print("Matrix dimension: {}".format(result['kernel_matrix_training'].shape))

print("testing success ratio: ", result['testing_accuracy'])

test_set = np.concatenate((test_input['Benign'], test_input['Malignant']))
y_test = qsvm.predict(test_set, quantum_instance)

##%% And here we plot the results. The first plot shows the label predictions of the QSVM and the second plot shows the test labels.
plt.scatter(test_set[:, 0], test_set[:,1], c=y_test)
plt.show()

plt.scatter(test_input['Benign'][:,0], test_input['Benign'][:,1])
plt.scatter(test_input['Malignant'][:,0], test_input['Malignant'][:,1])
plt.show()

$\endgroup$
2
$\begingroup$

I have previously used this function to load a custom data set - it should still work but I haven't tried it with more recent releases of Aqua

def userDefinedData(location, file, class_labels,training_size, test_size, n=2, PLOT_DATA=True):


    data, target, target_names = load_data(location, file)
    # sample_train is of the same form as data
    sample_train, sample_test, label_train, label_test = train_test_split(
        data, target,test_size=0.25, train_size=0.75 ,random_state=22)


    # Now we standarize for gaussian around 0 with unit variance
    std_scale = StandardScaler().fit(sample_train)
    sample_train = std_scale.transform(sample_train)
    sample_test = std_scale.transform(sample_test)


    # Now reduce number of features to number of qubits
    pca = PCA(n_components=n).fit(sample_train)
    sample_train = pca.transform(sample_train)
    sample_test = pca.transform(sample_test)


    # Samples are pairs of points
    samples = np.append(sample_train, sample_test, axis=0)
    minmax_scale = MinMaxScaler((-1, 1)).fit(samples)
    sample_train = minmax_scale.transform(sample_train)
    sample_test = minmax_scale.transform(sample_test)


    # If class labels are numeric
    if class_labels[0].isdigit():
        # Pick training size number of samples from each distro
        training_input = {key: (sample_train[label_train == int(key), :])[:training_size] for k, key in enumerate(class_labels)}
        test_input = {key: (sample_test[label_test == int(key), :])[: test_size] for k, key in enumerate(class_labels)}

    else:
        # if they aren't
        training_input = {key: (sample_train[label_train == k, :])[:training_size] for k, key in
                         enumerate(class_labels)}
        test_input = {key: (sample_train[label_train == k, :])[training_size:(
                training_size + test_size)] for k, key in enumerate(class_labels)}



    if PLOT_DATA:
        for k in range(0, 9):
            plt.scatter(sample_train[label_train == k, 0][:training_size],
                        sample_train[label_train == k, 1][:training_size])

        plt.title("PCA dim. reduced user dataset")
        plt.show()

    return sample_train, training_input, test_input, class_labels

| improve this answer | |
$\endgroup$
  • $\begingroup$ Let us continue this discussion in chat. $\endgroup$ – vis555 Feb 25 at 17:08
  • 1
    $\begingroup$ If you have further questions, it would be better to ask them as new questions rather than comments :) $\endgroup$ – met927 Feb 25 at 19:17
  • $\begingroup$ For any further discussion, please continue in the chatroom vis555 already set up. Thanks! $\endgroup$ – heather Feb 27 at 15:09

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.