<?xml version="1.0" encoding="utf-8" ?><rss version="2.0"><channel><title><![CDATA[机语]]></title><description><![CDATA[数学、机器学习与未来]]></description><link>https://blog.youkuaiyun.com/oidahdsah0</link><language>zh-cn</language><generator>https://blog.youkuaiyun.com/</generator><copyright><![CDATA[Copyright &copy; oidahdsah0]]></copyright><item><title><![CDATA[[ 娱乐算法 ] 位面之子算法 - The Chosen One Algorithm - Matlab code]]></title><link>https://blog.youkuaiyun.com/oidahdsah0/article/details/118964173</link><guid>https://blog.youkuaiyun.com/oidahdsah0/article/details/118964173</guid><author>oidahdsah0</author><pubDate>Wed, 21 Jul 2021 13:32:18 +0800</pubDate><description><![CDATA[位面之子算法 - The Chosen One Algorithm

最近书荒，想找点小说看，却发现没有什么能够耐心看下去的。
想要自己去写小说，却又发现动笔写大纲，塑造人物，建立剧情的前后关系，又是极其麻烦。
写游戏嘛？嘛，算了算了，代码块太多，本喵太懒
怎么办！干脆直接Simulation好了！说干就干！咱们这就模拟一个小说里强者为尊的世界！
实现语言，当然是简单便捷的Matlab！

声明：本代码为本人提出，仅供娱乐之用，欢迎转载，但请保留头部签名协议，感谢！
首先，这个世界需要规定一些参数：
%%.]]></description><category></category></item><item><title><![CDATA[文献阅读21期：Attention mechanisms and deep learning for machine vision: A survey]]></title><link>https://blog.youkuaiyun.com/oidahdsah0/article/details/118024404</link><guid>https://blog.youkuaiyun.com/oidahdsah0/article/details/118024404</guid><author>oidahdsah0</author><pubDate>Fri, 18 Jun 2021 16:08:08 +0800</pubDate><description><![CDATA[[ 综述阅读·机器视觉 ] Attention mechanisms and deep learning for machine vision:
A survey of the state of the art

推荐理由：一篇较新的综述，主在介绍Attention和配套的深度学习机制在机器视觉领域的最新成果，作者把来龙去脉讲得较为明白，一篇好文

1.摘要&amp;简介

最近几年，NLP领域较火的用于上下文预测识别的Attention机制获得了学界很大的认可，这引起了很多其他领域学者的注意。其中博主认为]]></description><category></category></item><item><title><![CDATA[文献阅读20期：Transformer Transforms Salient Object Detection and Camouflaged Object Detection]]></title><link>https://blog.youkuaiyun.com/oidahdsah0/article/details/118000129</link><guid>https://blog.youkuaiyun.com/oidahdsah0/article/details/118000129</guid><author>oidahdsah0</author><pubDate>Fri, 18 Jun 2021 01:33:39 +0800</pubDate><description><![CDATA[[ 文献阅读 ]  Transformer Transforms Salient Object Detection and Camouflaged Object Detection [1]

表现SOTA！性能优于SCWS、JLDCF等网络，源自机器翻译的Transformer网络特别擅长在长序列中对远程依存关系进行建模。大量实验结果表明，Transformer网络可以转换显著性目标检测和伪装对象检测，从而为每个相关任务提供了新的基准。

1.Transformer Network
1.1.Transfor]]></description><category></category></item><item><title><![CDATA[文献阅读19期：SEMI-SUPERVISED CLASSIFICATION WITH GRAPH CONVOLUTIONAL NETWORKS]]></title><link>https://blog.youkuaiyun.com/oidahdsah0/article/details/117931993</link><guid>https://blog.youkuaiyun.com/oidahdsah0/article/details/117931993</guid><author>oidahdsah0</author><pubDate>Wed, 16 Jun 2021 20:16:34 +0800</pubDate><description><![CDATA[[ 文献阅读·图卷积网络 ] SEMI-SUPERVISED CLASSIFICATION WITH GRAPH CONVOLUTIONAL NETWORKS [1]

推荐理由：

1.摘要&amp;简介
参考文献
[1] Kipf T N, Welling M. Semi-supervised classification with graph convolutional networks[J]. arXiv preprint arXiv:1609.02907, 2016.

...]]></description><category></category></item><item><title><![CDATA[文献阅读18期：Learning Convolutional Neural Networks for Graphs]]></title><link>https://blog.youkuaiyun.com/oidahdsah0/article/details/117902536</link><guid>https://blog.youkuaiyun.com/oidahdsah0/article/details/117902536</guid><author>oidahdsah0</author><pubDate>Mon, 14 Jun 2021 16:57:20 +0800</pubDate><description><![CDATA[[ 文献阅读·图神经网络 ] Learning Convolutional Neural Networks for Graphs [1]

推荐理由：本文考虑两个问题，一是训练一个可以用于未知图的函数，让未知图可以有效进行分类和回归问题求解。二是用于对未知图的特性进行预测（如点或边的缺失）。

1.摘要&amp;简介

本文的目的在于建立一种用于表达有向图或无向图的框架。
如果要理解图神经网络的卷积，那就得先理解传统的卷积神经网络CNN。如图1所示，从图卷积的角度去看，CNN就像是一个遍历所有节点，然后固定]]></description><category></category></item><item><title><![CDATA[文献阅读17期：Deep Learning on Graphs: A Survey - 6]]></title><link>https://blog.youkuaiyun.com/oidahdsah0/article/details/117786251</link><guid>https://blog.youkuaiyun.com/oidahdsah0/article/details/117786251</guid><author>oidahdsah0</author><pubDate>Thu, 10 Jun 2021 17:16:24 +0800</pubDate><description><![CDATA[[ 文献阅读·综述 ] Deep Learning on Graphs: A Survey [1]

推荐理由：图神经网络的survey paper，在很多的领域展现出了独特的作用力，分别通过GRAPH RNN（图循环网络）、GCN（图卷积）、GRAPH AUTOENCODERS（图自编码器）、GRAPH REINFORCEMENT LEARNING（图强化学习模型）、GRAPH ADVERSARIAL METHODS（图对抗模型）等五个类型的模型进行阐述，可以让大家对图神经网络有一个整体的认识。

6.图]]></description><category></category></item><item><title><![CDATA[文献阅读16期：Deep Learning on Graphs: A Survey - 5]]></title><link>https://blog.youkuaiyun.com/oidahdsah0/article/details/117669281</link><guid>https://blog.youkuaiyun.com/oidahdsah0/article/details/117669281</guid><author>oidahdsah0</author><pubDate>Tue, 08 Jun 2021 14:22:03 +0800</pubDate><description><![CDATA[[ 文献阅读·综述 ] Deep Learning on Graphs: A Survey [1]

推荐理由：图神经网络的survey paper，在很多的领域展现出了独特的作用力，分别通过GRAPH RNN（图循环网络）、GCN（图卷积）、GRAPH AUTOENCODERS（图自编码器）、GRAPH REINFORCEMENT LEARNING（图强化学习模型）、GRAPH ADVERSARIAL METHODS（图对抗模型）等五个类型的模型进行阐述，可以让大家对图神经网络有一个整体的认识。

5.图]]></description><category></category></item><item><title><![CDATA[文献阅读15期：Deep Learning on Graphs: A Survey - 4]]></title><link>https://blog.youkuaiyun.com/oidahdsah0/article/details/117630072</link><guid>https://blog.youkuaiyun.com/oidahdsah0/article/details/117630072</guid><author>oidahdsah0</author><pubDate>Mon, 07 Jun 2021 17:16:33 +0800</pubDate><description><![CDATA[[ 文献阅读·综述 ] Deep Learning on Graphs: A Survey [1]

推荐理由：图神经网络的survey paper，在很多的领域展现出了独特的作用力，分别通过GRAPH RNN（图循环网络）、GCN（图卷积）、GRAPH AUTOENCODERS（图自编码器）、GRAPH REINFORCEMENT LEARNING（图强化学习模型）、GRAPH ADVERSARIAL METHODS（图对抗模型）等五个类型的模型进行阐述，可以让大家对图神经网络有一个整体的认识。

4.3]]></description><category></category></item><item><title><![CDATA[文献阅读14期：Deep Learning on Graphs: A Survey - 3]]></title><link>https://blog.youkuaiyun.com/oidahdsah0/article/details/117600381</link><guid>https://blog.youkuaiyun.com/oidahdsah0/article/details/117600381</guid><author>oidahdsah0</author><pubDate>Sun, 06 Jun 2021 00:42:02 +0800</pubDate><description><![CDATA[[ 文献阅读·综述 ] Deep Learning on Graphs: A Survey [1]

推荐理由：图神经网络的survey paper，在很多的领域展现出了独特的作用力，分别通过GRAPH RNN（图循环网络）、GCN（图卷积）、GRAPH AUTOENCODERS（图自编码器）、GRAPH REINFORCEMENT LEARNING（图强化学习模型）、GRAPH ADVERSARIAL METHODS（图对抗模型）等五个类型的模型进行阐述，可以让大家对图神经网络有一个整体的认识。

4.2]]></description><category></category></item><item><title><![CDATA[文献阅读13期：Deep Learning on Graphs: A Survey - 2]]></title><link>https://blog.youkuaiyun.com/oidahdsah0/article/details/117567811</link><guid>https://blog.youkuaiyun.com/oidahdsah0/article/details/117567811</guid><author>oidahdsah0</author><pubDate>Sat, 05 Jun 2021 02:36:19 +0800</pubDate><description><![CDATA[[ 文献阅读·综述 ] Deep Learning on Graphs: A Survey [1]

推荐理由：图神经网络的survey paper，在很多的领域展现出了独特的作用力，分别通过GRAPH RNN（图循环网络）、GCN（图卷积）、GRAPH AUTOENCODERS（图自编码器）、GRAPH REINFORCEMENT LEARNING（图强化学习模型）、GRAPH ADVERSARIAL METHODS（图对抗模型）等五个类型的模型进行阐述，可以让大家对图神经网络有一个整体的认识

4. 图]]></description><category></category></item><item><title><![CDATA[文献阅读12期：Deep Learning on Graphs: A Survey - 1]]></title><link>https://blog.youkuaiyun.com/oidahdsah0/article/details/117515890</link><guid>https://blog.youkuaiyun.com/oidahdsah0/article/details/117515890</guid><author>oidahdsah0</author><pubDate>Fri, 04 Jun 2021 10:00:36 +0800</pubDate><description><![CDATA[[ 文献阅读·综述 ] Deep Learning on Graphs: A Survey [1]

推荐理由：图神经网络的survey paper，在很多的领域展现出了独特的作用力，分别通过GRAPH RNN（图循环网络）、GCN（图卷积）、GRAPH AUTOENCODERS（图自编码器）、GRAPH REINFORCEMENT LEARNING（图强化学习模型）、GRAPH ADVERSARIAL METHODS（图对抗模型）等五个类型的模型进行阐述，可以让大家对图神经网络有一个整体的认识

1.I.]]></description><category></category></item><item><title><![CDATA[文献阅读11期：A Deep Value-network Based Approach for Multi-Driver Order Dispatching]]></title><link>https://blog.youkuaiyun.com/oidahdsah0/article/details/117469859</link><guid>https://blog.youkuaiyun.com/oidahdsah0/article/details/117469859</guid><author>oidahdsah0</author><pubDate>Wed, 02 Jun 2021 23:52:52 +0800</pubDate><description><![CDATA[[ 文献阅读 ] A Deep Value-network Based Approach for Multi-Driver Order Dispatching [1]

推荐理由：滴滴Oral论文，新提出了一种新的基于深度强化学习与半马尔科夫决策过程的智能派单应用，在同时考虑时间与空间的长期优化目标的基础上利用深度神经网络进行更准确有效的价值估计。通过系统的离线模拟实验以及在滴滴平台的在线AB实验证明，这种基于深度强化学习的派单算法相比现有最好的方法能进一步显著提升平台各项效率及用户体验。

参考文献
[1]]></description><category></category></item><item><title><![CDATA[文献阅读10期：ATTENTION, LEARN TO SOLVE ROUTING PROBLEMS!]]></title><link>https://blog.youkuaiyun.com/oidahdsah0/article/details/117449584</link><guid>https://blog.youkuaiyun.com/oidahdsah0/article/details/117449584</guid><author>oidahdsah0</author><pubDate>Wed, 02 Jun 2021 00:10:25 +0800</pubDate><description><![CDATA[[ 文献阅读·路径规划 ]  ATTENTION, LEARN TO SOLVE ROUTING PROBLEMS! [1]

推荐理由：这篇应该不用多说了，ATTENTION模型做路径规划，算是一篇MileStone了。

1. Attention Model

一个TSP实例假定有n个节点，i∈{1,…,n}i \in\{1, \ldots, n\}i∈{1,…,n}，每个节点的特征为xi\mathbf{x}_{i}xi​，对于TSP问题来说，xi\mathbf{x}_{i}xi​即节点i的坐标，且构.]]></description><category></category></item><item><title><![CDATA[文献阅读09期：基于ADMM方法的柔性负载实时定价]]></title><link>https://blog.youkuaiyun.com/oidahdsah0/article/details/117415053</link><guid>https://blog.youkuaiyun.com/oidahdsah0/article/details/117415053</guid><author>oidahdsah0</author><pubDate>Mon, 31 May 2021 18:06:49 +0800</pubDate><description><![CDATA[[ 文献阅读·能源 ]  Load shifting of a supplier-based demand response of multi-class subscribers in smart grid [1]

推荐理由：

1.摘要&amp;简介
2.
3.
参考文献
[1] Li J, Liu B, Sun Q, et al. Load shifting of a supplier-based demand response of multi-class subscribers in smart ]]></description><category></category></item><item><title><![CDATA[文献阅读08期：SBM-DEA混合蒙特卡洛的能用效用评估]]></title><link>https://blog.youkuaiyun.com/oidahdsah0/article/details/117398698</link><guid>https://blog.youkuaiyun.com/oidahdsah0/article/details/117398698</guid><author>oidahdsah0</author><pubDate>Sun, 30 May 2021 17:38:47 +0800</pubDate><description><![CDATA[[ 文献阅读·能源 ] Energy supply efficiency evaluation of integrated energy systems using novel SBM-DEA integrating Monte Carlo [1]

推荐理由： 以往的DEA模型分为两种，径向模型和角度量模型，这两种模型要么没有考虑松弛问题，要么仅考虑单一的输入输出。本文提出的MC-SBM-DEA可以说是兼顾了两个方面，值得借鉴

1.摘要&amp;简介

主流的DEA模型分为两种，径向模型和角度量模型。.]]></description><category></category></item><item><title><![CDATA[文献阅读07期：智网中动态电价对太阳能板普及的影响]]></title><link>https://blog.youkuaiyun.com/oidahdsah0/article/details/117354726</link><guid>https://blog.youkuaiyun.com/oidahdsah0/article/details/117354726</guid><author>oidahdsah0</author><pubDate>Sat, 29 May 2021 15:46:56 +0800</pubDate><description><![CDATA[[ 文献阅读·能源 ] Techno-economic analysis of the impact of dynamic electricity prices on solar penetration in a smart grid environment with distributed energy storage

推荐理由：1.用斯塔克尔伯格博弈削峰填谷；2.


]]></description><category></category></item><item><title><![CDATA[文献阅读06期：基于Logistic效用函数智能电网定价]]></title><link>https://blog.youkuaiyun.com/oidahdsah0/article/details/117326005</link><guid>https://blog.youkuaiyun.com/oidahdsah0/article/details/117326005</guid><author>oidahdsah0</author><pubDate>Thu, 27 May 2021 15:28:12 +0800</pubDate><description><![CDATA[[ 文献阅读·能源 ] The real-time pricing optimization model of smart grid based on the utility function of the logistic function [1]

推荐理由：需求侧响应（需求侧管理）与社会福利最大化最近几年在能源领域逐渐受到重视，本文提出一种基于Logistic效用函数

1.摘要&amp;简介

效用函数是求解智能电网实时电价的一个非常关键的因素。本文根据Logistic函数提出一种新型效用函数。
问]]></description><category></category></item><item><title><![CDATA[文献阅读05期：GLaRA - 弱监督实体识别的图方法扩充]]></title><link>https://blog.youkuaiyun.com/oidahdsah0/article/details/117298761</link><guid>https://blog.youkuaiyun.com/oidahdsah0/article/details/117298761</guid><author>oidahdsah0</author><pubDate>Wed, 26 May 2021 21:54:36 +0800</pubDate><description><![CDATA[[ 文献阅读·DL ] Outline：GLARA: Graph-based Labeling Rule Augmentation forWeakly Supervised Named Entity Recognition

推荐理由：


]]></description><category></category></item><item><title><![CDATA[文献阅读04期：神经网络解释 - GRACE]]></title><link>https://blog.youkuaiyun.com/oidahdsah0/article/details/117253363</link><guid>https://blog.youkuaiyun.com/oidahdsah0/article/details/117253363</guid><author>oidahdsah0</author><pubDate>Wed, 26 May 2021 03:29:28 +0800</pubDate><description><![CDATA[[ 文献阅读·DL ] Outline：GRACE: Generating Concise and Informative Contrastive Sample to Explain Neural Network Model’s Prediction

推荐理由：虽然神经网络近年来越来越为人所熟知，但是网络的可解释性一直是一个值得关注的话题，而且想要这个领域作出一定成果，属实一项挑战。该论文借鉴了因果关系中的“干预解释”和哲学中的“解释性是对比性”，从而提出了一种名为GRACE的新颖解决方案，该解决方案可以]]></description><category></category></item><item><title><![CDATA[文献阅读03期：A bilevel programming method 一种针对包含不确定性微网实时定价的双层优化方法]]></title><link>https://blog.youkuaiyun.com/oidahdsah0/article/details/117208251</link><guid>https://blog.youkuaiyun.com/oidahdsah0/article/details/117208251</guid><author>oidahdsah0</author><pubDate>Mon, 24 May 2021 16:43:21 +0800</pubDate><description><![CDATA[[ 文献阅读 ] Outline：Real-time pricing for smart grid with multi-energy microgrids and uncertain loads: a bilevel programming method

]]></description><category></category></item></channel></rss>