当前位置：首页 > news >正文

deeplog中输出某个 event 的概率

news 2025/9/4 13:54:02

1 实现之后效果

# import DeepLog and Preprocessor
import numpy as np
from deeplog import DeepLog
import torch# Create DeepLog object
deeplog = DeepLog(input_size  = 10, # Number of different events to expecthidden_size = 64 , # Hidden dimension, we suggest 64output_size = 10, # Number of different events to expect
)# X数据维度 30×10
X = torch.randint(1,8, size=(30, 10))
# 标签
Y = np.random.randint(1,8, size=30)
# 输出每个标签的概率
result = deeplog.predict_prob(X = X,y = Y)print(result.shape)
print(result)

输出结果：
在这里插入图片描述

2 实现步骤

step1 找到安装包位置，并打开文件
在这里插入图片描述

step2 DeepLog 类中添加如下函数

class DeepLog(Module):..................def predict_prob(self, X, y, k=1, variable=False, verbose=True):"""Predict the k most likely output valuesParameters----------X : torch.Tensor of shape=(n_samples, seq_len)Input of sequences, these will be one-hot encoded to an array ofshape=(n_samples, seq_len, input_size)y : IgnoredIgnoredk : int, default=1Number of output items to generatevariable : boolean, default=FalseIf True, predict inputs of different sequence lengthsverbose : boolean, default=TrueIf True, print outputReturns-------result : torch.Tensor of shape=(n_samples, k)k most likely outputsconfidence : torch.Tensor of shape=(n_samples, k)Confidence levels for each output"""# Get the predictionsresult = super().predict(X, variable=variable, verbose=verbose)# Get the probabilities from the log probabilitiesresult = result.exp()# return a given key's probindex_c = yindex_r = torch.arange(y.shape[0])return result[index_r, index_c]