Quantum SVM via SOCP experiment logs

The dataset contains the experiment logs for the "Quantum algorithms for Second-Order Cone Programming and Support Vector Machines" paper. Every row represents the details of solving a single random SVM instance. The meanings of the columns are the following:n, m: dimension and number of training examplescorruption: the label corruption probability p from the paperacc_train, acc_test: classification accuracies achieved by the simulated quantum algorithmacc_socp_train, acc_socp_test: idem, but for a SVM trained by solving the SOCP reduction with ECOSsocp_train_time: training time for ECOS, in secondslibsvm_train_time: training time for LIBSVM with a linear kernel<br>matrix_size: size of the Newton systemmu: the final duality gaplambda_min_x, lambda_min_s: final distances from the cone boundarycp_dist: final distance from the central pathkappa: final condition number of the Newton matrixfrob: the zeta parameter, ||A||_F/||A||_2 for the Newton matrixdifficulty: the data dependent factor in the complexity expression (arbitrary units)<br>complexity: the product of all quantities in the complexity expression (arbitrary units)exit_reason: useful for debugging, indicates which convergence criterion was triggered<br><br>