calcParam.py

import struct
import numpy as np
from sklearn.model_selection import GridSearchCV
from sklearn.svm import SVC


def parse(filename):
    """Parse data file created by getdata"""
    with open(filename, "r") as f:
        raw_data = [int(line.split()[1], 16) for line in f]

    signal = [[], [], [], []]
    signal_data = [[], [], [], []]
    integ_data = [[], [], [], []]
    svm_data = [[], [], [], []]
    sort_data = [[], [], [], []]

    for x in raw_data:
        dtype = x >> 30
        ch = (x >> 28) & 3
        if dtype == 0:
            if (x >> 27) & 1:
                signal[ch] = []

            signal[ch].append(x & 0x3fff)

            if (x >> 26) & 1:
                signal_data[ch].append(signal[ch])

        elif dtype == 1:
            integ_data[ch].append(x & 0x3ffffff)

        elif dtype == 2:
            fl = struct.unpack('f', struct.pack('I', (x << 4) & 0xffffffff))[0]
            svm_data[ch].append(fl)

        elif dtype == 3:
            sort_data[ch].append(x & 0x1ffffff)

    return signal_data, integ_data, svm_data, sort_data


def normalize(X):
    mu = X.mean(axis=0)
    sigma = X.std(axis=0)
    return (X-mu)/sigma, mu, sigma


def train(pos, neg):
    pos = np.array(pos)
    neg = np.array(neg)
    assert len(pos.shape) == 2 and len(neg.shape) == 2
    assert pos.shape[1] == neg.shape[1]

    param_grid = {'C': [0.01, 0.02, 0.05, 0.1, 0.2, 0.5, 1, 2, 5, 10, 20, 50,
                        100, 200, 500, 1000, 2000, 5000]}
    model = GridSearchCV(SVC(), param_grid=param_grid, n_jobs=-1,
                         cv=min(10, min(len(pos), len(neg))))
    X = np.concatenate([pos, neg])
    y = np.array([1] * len(pos) + [0] * len(neg))
    model.fit(X, y)
    return model.best_estimator_


def write(estimator, mu, sigma):
    alpha = estimator.dual_coef_
    if type(estimator.gamma) == float:
        K = estimator.gamma
    elif estimator.gamma == 'auto':
        K = estimator.support_vectors_.shape[1]
    else:
        raise Exception('unknown gamma parameter')

    b = estimator.intercept_

    np.savetxt('alpah.csv', alpha.reshape(1, -1), delimiter=',')
    np.savetxt('k.csv', K.reshape(1, -1), delimiter=',')
    np.savetxt('b.csv', b.reshape(1, -1), delimiter=',')
    np.savetxt('mu.csv', mu.reshape(1, -1), delimiter=',')
    np.savetxt('sigma.csv', sigma.reshape(1, -1), delimiter=',')