Exemplos de loguniform em Python

Exemplo n.º 1

Arquivo: tune.py Projeto: nickcorona/india_housing

    num_boost_round=MAX_ROUNDS,
    early_stopping_rounds=EARLY_STOPPING_ROUNDS,
    verbose_eval=REPORT_ROUNDS,
)

lgb.plot_importance(model,
                    grid=False,
                    max_num_features=20,
                    importance_type="gain")
plt.show()

TUNE_ETA = True
best_etas = {"learning_rate": [], "score": []}
if TUNE_ETA:
    for _ in range(120):
        eta = loguniform(-5, 1)
        best_etas["learning_rate"].append(eta)
        params["learning_rate"] = eta
        model = lgb.train(
            params,
            dt,
            valid_sets=[dt, dv],
            valid_names=["training", "valid"],
            num_boost_round=MAX_ROUNDS,
            early_stopping_rounds=EARLY_STOPPING_ROUNDS,
            verbose_eval=False,
        )
        best_etas["score"].append(model.best_score["valid"][METRIC])

    best_eta_df = pd.DataFrame.from_dict(best_etas)
    lowess_data = lowess(

Exemplo n.º 2

    params,
    dt,
    valid_sets=[dt, dv],
    valid_names=["training", "valid"],
    num_boost_round=MAX_ROUNDS,
    early_stopping_rounds=EARLY_STOPPING_ROUNDS,
    verbose_eval=REPORT_ROUNDS,
)

lgb.plot_importance(model, grid=False, importance_type="gain")
plt.show()

best_etas = {"learning_rate": [], "score": []}

for _ in range(200):
    eta = loguniform(-3, 0)
    best_etas["learning_rate"].append(eta)
    params["learning_rate"] = eta
    model = lgb.train(
        params,
        dt,
        valid_sets=[ds, dv],
        valid_names=["training", "valid"],
        num_boost_round=MAX_ROUNDS,
        early_stopping_rounds=EARLY_STOPPING_ROUNDS,
        verbose_eval=False,
    )
    best_etas["score"].append(model.best_score["valid"][METRIC])

best_eta_df = pd.DataFrame.from_dict(best_etas)
lowess_data = lowess(

Exemplo n.º 3

    params,
    dt,
    valid_sets=[dt, dv],
    valid_names=["training", "valid"],
    num_boost_round=MAX_ROUNDS,
    early_stopping_rounds=EARLY_STOPPING_ROUNDS,
    verbose_eval=REPORT_ROUNDS,
)

lgb.plot_importance(model, grid=False, importance_type="gain")
plt.show()

best_etas = {"learning_rate": [], "score": []}

for _ in range(30):
    eta = loguniform(-1, 0)
    best_etas["learning_rate"].append(eta)
    params["learning_rate"] = eta
    model = lgb.train(
        params,
        dt,
        valid_sets=[ds, dv],
        valid_names=["training", "valid"],
        num_boost_round=MAX_ROUNDS,
        early_stopping_rounds=EARLY_STOPPING_ROUNDS,
        verbose_eval=False,
    )
    best_etas["score"].append(model.best_score["valid"][METRIC])

best_eta_df = pd.DataFrame.from_dict(best_etas)
lowess_data = lowess(

Exemplo n.º 4

Arquivo: tune.py Projeto: nickcorona/greyhound_racing

    valid_names=["training", "valid"],
    num_boost_round=MAX_ROUNDS,
    early_stopping_rounds=EARLY_STOPPING_ROUNDS,
    verbose_eval=REPORT_ROUNDS,
)

lgb.plot_importance(model,
                    grid=False,
                    max_num_features=20,
                    importance_type="gain")
plt.show()

best_etas = {"learning_rate": [], "score": []}

for _ in range(60):
    eta = loguniform(-4, 0)
    best_etas["learning_rate"].append(eta)
    params["learning_rate"] = eta
    model = lgb.train(
        params,
        dt,
        valid_sets=[dt, dv],
        valid_names=["training", "valid"],
        num_boost_round=MAX_ROUNDS,
        early_stopping_rounds=EARLY_STOPPING_ROUNDS,
        verbose_eval=False,
    )
    best_etas["score"].append(model.best_score["valid"][METRIC])

best_eta_df = pd.DataFrame.from_dict(best_etas)
lowess_data = lowess(

Exemplo n.º 5

    num_boost_round=MAX_ROUNDS,
    early_stopping_rounds=EARLY_STOPPING_ROUNDS,
    verbose_eval=REPORT_ROUNDS,
)

lgb.plot_importance(model,
                    grid=False,
                    max_num_features=20,
                    importance_type="gain")
plt.show()

TUNE_ETA = True
best_etas = {"learning_rate": [], "score": []}
if TUNE_ETA:
    for _ in range(30):
        eta = loguniform(-6, 0)
        best_etas["learning_rate"].append(eta)
        params["learning_rate"] = eta
        model = lgb.train(
            params,
            dt,
            valid_sets=[dt, dv],
            valid_names=["training", "valid"],
            num_boost_round=MAX_ROUNDS,
            early_stopping_rounds=EARLY_STOPPING_ROUNDS,
            verbose_eval=False,
        )
        best_etas["score"].append(model.best_score["valid"][METRIC])

    best_eta_df = pd.DataFrame.from_dict(best_etas)
    lowess_data = lowess(