hmc_loss 1.0.0

结构

python实现的分层多类验证度量：hmc loss . 原纸是(Bi&Kwok, 2012)。

安装

pip install hmc_loss

要求

• 努比
• 网络x

如何使用

这个度量是像scikit学习度量一样实现的。

from hmc_loss import hmc_loss_score, get_cost_list
import numpy as np

# Generate label data(2-D array of numpy)
true_label = np.random.randint(2, size(100, 100))
pred_label = np.random.randint(2, size(100, 100))

# Generate test graph(Di-Graph of NetworkX)
graph = nx.gnc_graph(100)
# Generate element list of graph node
label_list = list(range(100))
# Calculate cost of each node in graph
cost_list = get_cost_list(graph, 0, label_list)
# Calculate HMC-loss
hmc_loss_score(true_label, pred_label, graph, 0, label_list, cost_list, alpha=0.5, beta=1.5)

许可证

MIT

作者

Taske HAMANO