<?xml version="1.0" encoding="utf-8" ?><rss version="2.0"><channel><title><![CDATA[Cyrus_May的博客]]></title><description><![CDATA[]]></description><link>https://blog.csdn.net/Cyrus_May</link><language>zh-cn</language><generator>https://blog.csdn.net/</generator><copyright><![CDATA[Copyright &copy; Cyrus_May]]></copyright><item><title><![CDATA[BERT网络的原理与实战]]></title><link>https://blog.csdn.net/Cyrus_May/article/details/130856290</link><guid>https://blog.csdn.net/Cyrus_May/article/details/130856290</guid><author>Cyrus_May</author><pubDate>Wed, 24 May 2023 21:50:14 +0800</pubDate><description><![CDATA[BERT（Bidirectional Encoder Representations from Transformers）是一种基于Transformer架构的预训练语言模型，由Google在2018年提出。BERT可以在大规模的未标注文本上进行预训练，然后在各种下游NLP任务上进行微调，取得了很好的效果。BERT的主要贡献在于将双向预训练引入了Transformer架构中，使得模型能够更好地理解上下文信息，从而在下游任务中表现更加出色。本文将介绍BERT网络的原理与实战，包括预训练和微调两个部分。]]></description><category></category></item><item><title><![CDATA[Transformer网络原理与实战]]></title><link>https://blog.csdn.net/Cyrus_May/article/details/130835430</link><guid>https://blog.csdn.net/Cyrus_May/article/details/130835430</guid><author>Cyrus_May</author><pubDate>Tue, 23 May 2023 21:07:58 +0800</pubDate><description><![CDATA[Transformer网络是一种基于自注意力机制的神经网络，由Google于2017年提出，并被广泛应用于自然语言处理、语音识别、图像生成等领域。相对于传统的循环神经网络和卷积神经网络，Transformer网络具有更好的并行性和更高的计算效率，在处理长文本时表现更加出色。Transformer网络的核心思想是利用自注意力机制来实现序列建模。该模型的输入和输出都是序列，它可以将源序列和目标序列分别映射到一个连续的向量空间中，并使用自注意力机制来计算输入序列中各个位置之间的依赖关系，从而实现对序列的建模。]]></description><category></category></item><item><title><![CDATA[XGBoost算法原理与实战]]></title><link>https://blog.csdn.net/Cyrus_May/article/details/130817299</link><guid>https://blog.csdn.net/Cyrus_May/article/details/130817299</guid><author>Cyrus_May</author><pubDate>Mon, 22 May 2023 23:16:53 +0800</pubDate><description><![CDATA[XGBoost全名叫做eXtreme Gradient Boosting，是一种基于GBDT的高效、灵活、可扩展的梯度提升算法。近年来，在各种数据科学竞赛中，XGBoost屡创佳绩，成为众多数据科学家和算法工程师的新宠。本文将从原理和实战两个方面，带领大家深入了解XGBoost。]]></description><category></category></item><item><title><![CDATA[Soft Actor-Critic(SAC算法)]]></title><link>https://blog.csdn.net/Cyrus_May/article/details/126714184</link><guid>https://blog.csdn.net/Cyrus_May/article/details/126714184</guid><author>Cyrus_May</author><pubDate>Tue, 06 Sep 2022 20:34:05 +0800</pubDate><description><![CDATA[经过不断地soft policy evaluation和policy improvement，最终policy会收敛至。当|A|]]></description><category></category></item><item><title><![CDATA[强化学习—— Twin delay deep deterministic policy gradient(TD3算法)]]></title><link>https://blog.csdn.net/Cyrus_May/article/details/126492788</link><guid>https://blog.csdn.net/Cyrus_May/article/details/126492788</guid><author>Cyrus_May</author><pubDate>Tue, 23 Aug 2022 21:57:00 +0800</pubDate><description><![CDATA[所以动作价值的估计函数学习的目标是累计回报与TD error之差的期望。--------	从replay buffer中随机采样一个batch。使用了两个动作价值网络和一个策略网络，对应于三个Target 网络。----------------	更新Target network。--------将transition。初始化replay buffer。初始化Target网络中的参数。--------更新价值网络。，并随机初始化其中的参数。--------得到奖励。，并得到下一时刻的状态。]]></description><category></category></item><item><title><![CDATA[linux - vi / vim 编辑器]]></title><link>https://blog.csdn.net/Cyrus_May/article/details/126355494</link><guid>https://blog.csdn.net/Cyrus_May/article/details/126355494</guid><author>Cyrus_May</author><pubDate>Fri, 19 Aug 2022 21:34:10 +0800</pubDate><description><![CDATA[退出编辑1模式 ESC。]]></description><category></category></item><item><title><![CDATA[Linux（一）—— 文件系统]]></title><link>https://blog.csdn.net/Cyrus_May/article/details/126238171</link><guid>https://blog.csdn.net/Cyrus_May/article/details/126238171</guid><author>Cyrus_May</author><pubDate>Mon, 08 Aug 2022 23:35:23 +0800</pubDate><description><![CDATA[用户所需要的所有应用程序和数据（用户级的东西）]]></description><category></category></item><item><title><![CDATA[Pytorch —— 分布式模型训练]]></title><link>https://blog.csdn.net/Cyrus_May/article/details/126063176</link><guid>https://blog.csdn.net/Cyrus_May/article/details/126063176</guid><author>Cyrus_May</author><pubDate>Fri, 29 Jul 2022 21:43:28 +0800</pubDate><description><![CDATA[在每个节点上都执行一次。]]></description><category></category></item><item><title><![CDATA[强化学习—— Trust Region Policy Optimization (TRPO算法)]]></title><link>https://blog.csdn.net/Cyrus_May/article/details/125793435</link><guid>https://blog.csdn.net/Cyrus_May/article/details/125793435</guid><author>Cyrus_May</author><pubDate>Thu, 14 Jul 2022 21:34:25 +0800</pubDate><description><![CDATA[TRPO]]></description><category></category></item><item><title><![CDATA[Pytorch（六） —— 模型调优tricks]]></title><link>https://blog.csdn.net/Cyrus_May/article/details/125588021</link><guid>https://blog.csdn.net/Cyrus_May/article/details/125588021</guid><author>Cyrus_May</author><pubDate>Sun, 03 Jul 2022 20:51:14 +0800</pubDate><description><![CDATA[Pytorch]]></description><category></category></item><item><title><![CDATA[Ubuntu系统使用]]></title><link>https://blog.csdn.net/Cyrus_May/article/details/125585529</link><guid>https://blog.csdn.net/Cyrus_May/article/details/125585529</guid><author>Cyrus_May</author><pubDate>Sun, 03 Jul 2022 18:10:26 +0800</pubDate><description><![CDATA[Ubuntu使用]]></description><category></category></item><item><title><![CDATA[Pytorch（四）——数据加载及预处理]]></title><link>https://blog.csdn.net/Cyrus_May/article/details/125539321</link><guid>https://blog.csdn.net/Cyrus_May/article/details/125539321</guid><author>Cyrus_May</author><pubDate>Thu, 30 Jun 2022 17:13:48 +0800</pubDate><description><![CDATA[pytorch]]></description><category></category></item><item><title><![CDATA[Pytorch（四） —— 可视化工具 Visdom]]></title><link>https://blog.csdn.net/Cyrus_May/article/details/125520658</link><guid>https://blog.csdn.net/Cyrus_May/article/details/125520658</guid><author>Cyrus_May</author><pubDate>Wed, 29 Jun 2022 15:47:46 +0800</pubDate><description><![CDATA[pytorch]]></description><category></category></item><item><title><![CDATA[Pytorch（三） —— 函数优化]]></title><link>https://blog.csdn.net/Cyrus_May/article/details/125519826</link><guid>https://blog.csdn.net/Cyrus_May/article/details/125519826</guid><author>Cyrus_May</author><pubDate>Wed, 29 Jun 2022 13:52:10 +0800</pubDate><description><![CDATA[pytorch]]></description><category></category></item><item><title><![CDATA[Pytorch（二） —— 激活函数、损失函数及其梯度]]></title><link>https://blog.csdn.net/Cyrus_May/article/details/125500584</link><guid>https://blog.csdn.net/Cyrus_May/article/details/125500584</guid><author>Cyrus_May</author><pubDate>Tue, 28 Jun 2022 17:14:55 +0800</pubDate><description><![CDATA[δ(x)=11+e−xδ′(x)=δ(1−δ)\delta(x)=\frac{1}{1+e^{-x}}\\\delta'(x)=\delta(1-\delta)δ(x)=1+e−x1​δ′(x)=δ(1−δ)
tanh(x)=ex−e−xex+e−x∂tanh(x)∂x=1−tanh2(x)tanh(x)=\frac{e^x-e^{-x}}{e^x+e^{-x}}\\\frac{\partial tanh(x)}{\partial x}=1-tanh^2(x)tanh(x)=ex+e−xex−e−x​∂x∂]]></description><category></category></item><item><title><![CDATA[Pytorch（一） —— 基本语法]]></title><link>https://blog.csdn.net/Cyrus_May/article/details/125441729</link><guid>https://blog.csdn.net/Cyrus_May/article/details/125441729</guid><author>Cyrus_May</author><pubDate>Sat, 25 Jun 2022 17:55:21 +0800</pubDate><description><![CDATA[pytorch的基本语法和常用方法]]></description><category></category></item><item><title><![CDATA[Tensorflow（三十四） —— GRU]]></title><link>https://blog.csdn.net/Cyrus_May/article/details/124214692</link><guid>https://blog.csdn.net/Cyrus_May/article/details/124214692</guid><author>Cyrus_May</author><pubDate>Mon, 18 Apr 2022 11:33:46 +0800</pubDate><description><![CDATA[Tensorflow（三十四）  ——  GRU
import tensorflow as tf 
from tensorflow import keras
import numpy as np
tf.compat.v1.disable_eager_execution()

# 加载数据集
total_words = 10000
(x_train,y_train),(x_test,y_test) = keras.datasets.imdb.load_data(num_words=total_words)
p]]></description><category></category></item><item><title><![CDATA[Tensorflow（三十三） —— LSTM]]></title><link>https://blog.csdn.net/Cyrus_May/article/details/124214639</link><guid>https://blog.csdn.net/Cyrus_May/article/details/124214639</guid><author>Cyrus_May</author><pubDate>Mon, 18 Apr 2022 11:33:31 +0800</pubDate><description><![CDATA[Tensorflow（三十三）  ——  LSTMLSTM(Long Short-Term Memory)
LSTM(Long Short-Term Memory)

import tensorflow as tf
from tensorflow import keras
import matplotlib.pyplot as plt
import pandas as pd
from tensorflow.keras.preprocessing.sequence import pad_sequences

]]></description><category></category></item><item><title><![CDATA[Tensorflow（三十二） —— 梯度爆炸]]></title><link>https://blog.csdn.net/Cyrus_May/article/details/124214593</link><guid>https://blog.csdn.net/Cyrus_May/article/details/124214593</guid><author>Cyrus_May</author><pubDate>Mon, 18 Apr 2022 11:33:19 +0800</pubDate><description><![CDATA[Tensorflow（三十二）  ——  梯度爆炸解决方案
解决方案
import tensorflow as tf
from tensorflow import keras

# solve gradient exploding
with tf.GradientTape() as tape:
    logits = model(x)
    loss = tf.losses.categorical_crossentropy(y,logits,from_logits=True)
grads = tape.]]></description><category></category></item><item><title><![CDATA[Tensorflow（三十一） —— RNN]]></title><link>https://blog.csdn.net/Cyrus_May/article/details/124214497</link><guid>https://blog.csdn.net/Cyrus_May/article/details/124214497</guid><author>Cyrus_May</author><pubDate>Mon, 18 Apr 2022 11:33:00 +0800</pubDate><description><![CDATA[Tensorflow（三十一）  ——  RNN
import tensorflow as tf
from tensorflow import keras

# ***************** 1、input dim &amp; hidden dim
cell = keras.layers.SimpleRNNCell(64) # 4为隐含层神经元数量
cell.build(input_shape = [None,6])
print(cell.trainable_variables[0].shape,\
]]></description><category></category></item></channel></rss>