e2_config_hddm.py

"""
Config for additional experiment 2 including hddm
"""
import strlearn as sl
from sklearn.naive_bayes import GaussianNB
import numpy as np
from methods import DDM, EDDM, ADWIN, ALWAYS, NEVER, Meta, SDDM, SDDE, HDDM_AA, HDDM_WW

def e2_methods():
    return [
        GaussianNB(),
    ]

def e2_methods_labels():
    return [
        'GNB',
    ]

def metrics():
    return [
        sl.metrics.balanced_accuracy_score
        ]

def metrics_names():
    return [
        "BAC"
        ]

def e2_replications():
    return 10

def e2_static():
    return {
        'n_drifts': 5,
        'n_chunks': 200,
        'chunk_size': 250,
        'n_features': 15,
        'n_informative': 15,
        'n_redundant': 0
        }

def e2_static2():
    return {
        # 'n_drifts': 5,
        'n_chunks': 200,
        'chunk_size': 250,
        # 'n_features': 15,
        # 'n_informative': 15,
        'n_redundant': 0
        }

def e2_drift_types():
    return {
    'sudden': {},
    'gradual': {
    'concept_sigmoid_spacing': 5
        },
    'incremental': {
    'concept_sigmoid_spacing': 5,
    'incremental': True
        },
    }

def e2_recurring():
    return {
        'recurring': {
            'recurring': True
        },
        'not-recurring': {}
    }

def e2_clf_names():
    names = []
    detectors = ['HDDM_W', 'HDDM_A']

    for d in detectors:
        for m in e2_methods_labels():
            # names.append(d + "-" + m)
            names.append(d)

    return names

def e2_n_drifts():
    return {
        # 1: { 'n_drifts': 1},
        3: { 'n_drifts': 3},
        5: { 'n_drifts': 5},
        7: { 'n_drifts': 7},
        # 9: { 'n_drifts': 9},
    }

def e2_n_features():
    return {
        10: { 'n_features': 10, 'n_informative': 10},
        15: { 'n_features': 15, 'n_informative': 15},
        20: { 'n_features': 20, 'n_informative': 20},
        # 25: { 'n_features': 25, 'n_informative': 25},
        # 30: { 'n_features': 30, 'n_informative': 30},
    }

def e2_clfs(sdde_n_det=0, sdde_sensitivity=0):
    clfs = []

    # HDDM_W
    for m in e2_methods():
        clfs.append(Meta(detector = HDDM_WW(), base_clf = m))

    # HDDM_A
    for m in e2_methods():
        clfs.append(Meta(detector = HDDM_AA(), base_clf = m))



    return clfs