文献学习——A Deep Dive into Conflict Generating Decisions
A Deep Dive into Conflict Generating Decisions
- Md. Solimul Chowdhury, Martin Müller, Jia-Huai You:
A Deep Dive into Conflict Generating Decisions. CoRR abs/2105.04595 (2021) - Department of Computing Science, University of Alberta. 阿尔伯塔大学计算机科学系 {mdsolimu, mmueller, jyou}@ualberta.ca
Abstract. Boolean Satisfiability (SAT) is a well-known NP-complete problem. Despite this theoretical hardness, SAT solvers based on Conflict Driven Clause Learning (CDCL) can solve large SAT instances from many important domains. CDCL learns clauses from conflicts, a technique that allows a solver to prune its search space. The selection heuristics in CDCL prioritize variables that are involved in recent conflicts. While only a fraction of decisions generate any conflicts, many generate In this paper, we study conflict-generating decisions in CDCL in detail. We investigate the impact of single conflict (sc) decisions, which generate only one conflict, and multi-conflict (mc) decisions which generate two or more. We empirically characterize these two types of decisions based on the quality of the learned clauses produced by each type of decision. We also show an important connection between consecutive clauses learned within the same mc decision, where one learned clause triggers 译文:我们还展示了在同一个mc决策中学习的连续子句之间的重要联系,其中一个学习到的子句触发下一个子句的学习,形成一个子句链。 译文:这导致我们考虑冲突之间的相似性,为此我们制定了冲突邻近性的概念作为相似性度量。 译文:最后,我们提出了公共原因变量约简(CRVR)作为一种新的决策策略,从mc决策的学习子句中降低一些变量的选择优先级。 |
|
1 Introduction ...... The clause learning process in CDCL can generate more than one conflict for one decision. In the following, we categorize each conflict-producing decision as a single conflict (sc) or a multi-conflicts (mc) decision, depending on whether it produces one, or more than one, conflict. We label the resulting learned clauses sc and mc clauses accordingly. 译文:CDCL中的子句学习过程可以为一个决策产生多个冲突。在下面,我们将每个产生冲突的决策分类为单个冲突(sc)或多个冲突(mc)决策,这取决于它是产生一个还是多个冲突。我们将得到的学习到的从句相应地标记为sc和mc从句。 Conflicts play a crucial role in CDCL search. A better understanding of conflict generating decisions is a step towards a better understanding of CDCL and may open up new directions to improve CDCL search. Motivated by this, here we study conflict producing decisions in CDCL. The contributions of this work are: 译文:冲突在CDCL搜索中起着关键作用。更好地理解冲突生成决策是更好地理解CDCL的一步,并可能为改进CDCL搜索开辟新的方向。基于此,本文研究了CDCL中的冲突产生决策。这项工作的贡献是: |
|
2 Preliminaries Basic Operations of CDCL Conflicts and Clause Learning Relevant Notions Global Learning Rate The Literal Block Distance (LBD) Score Glue Clauses Glue to Learned 2.2 Notation sc and mc Decisions Burst of mc Decisions We define the burst of a mc decision as the number of conflicts (i.e., learned clauses) generated within that mc decision. Learned Clause Quality Over sc and mc Decisions 2.3 Test Set, Experimental Setup and Solvers Used |
|
3 An Empirical Analysis of sc and mc Decisions 3.1 Distributions of sc and mc decisions We denote Percentage of Decisions with Single Conflict and Percentage of Decisions with Multiple Conflicts as PDSC and PDMC, respectively. On average, about 21% (8%+13%) of the decisions are conflict producing. PDSC 8% and PDMC 13%。 However, since the mc decisions produce 2.65 (Column F) conflicts on average, this results in the generation of almost 1 conflict per 2 decisions, on average, which is reflected in the average GLR value of 0.49 for these instances. 3.2 Learned Clause Quality in sc and mc Decisions Fig. 1 shows per-instance details of these three measures in log scale. In almost all instances, LBD scores for mc To summarize, on average mc decisions are conflict-inefficient compared to sc decisions. However, on average the best quality learned clause from a mc decision has better quality than the quality of a sc clause. 3.3 Bursts of mc Decisions Burst of mc Decisions ——We define the burst of a mc decision as the number of conflicts (i.e., learned clauses) generated within that mc decision 在MC决策中产生的冲突(即学习的从句)的数量 Column F in Table 1 shows the average value of avgBurst for the test set. On average, the burst of mc decisions are quite small, about 2.65. However, as shown in column G, the average value of maxBurst is very high. The left plot in Fig. 2 compares these values for each test instance in log scale. This indicates that while large bursts of mc decisions occur, they are rare, as indicated by the average of 2.65. To analyze this in detail, we count the number of mc decisions for each burst size from 2 to 10. To analyze this in detail, we count the number of mc decisions for each burst size from 2 to 10. Distribution of mc Decisions by Burst Size —— The frequency of bursts decreases exponentially with their size. 爆发的频率随着它们的大小呈指数下降 |
|
4 Clause Learning in mc Decisions In this section, we establish a structural property of the learned clauses in mc decisions. 建立MC决策中所学从句的结构属性 注释:mc决策实际上是A、B两类连续冲突的B类。A类连续冲突是连续决策每次都导致生成冲突(前一次冲突找到的UIP文字在回溯后加入传播序列,紧接着BCP传播没有冲突,于是下一次决策,又发生冲突。)B类连续冲突是本文中一次决策形成冲突,由于冲突分析得到的UIP文字进入回溯后的传播队列和学习子句加入学习子句集,在随后的BCP是形成下一次冲突,并可能形成再一次冲突,从而形成连续的冲突图。本文献对B类连续冲突给出了较为形象的表示,也给出了科学规范的文字符号表述。了解这个情况,看本文的表述就会比较明白了,对照code修改也不难了。 |
|
5 Proximity between Conflicts Sequences in CDCL CDCL中冲突序列之间的接近性 introduce the measure of ConflictsProximity 5.1 Conflicts Proximity Literal Block Proximity 5.2 Proximity of Conflicts over sc and mc Decisions 我们现在在冲突邻近性下研究CDCL中的冲突邻近性。We now study proximity of conflicts in CDCL under ConflictsProximity. |
|
6 The Common Reason Variable Reduction Strategy 6.1 Common Reason Decision Variables a common reason decision level a common reason decision variable (CRV) 6.2 Poor mc Decisions 6.3 The CRVR Decision Strategy 抑制此类crv对未来决策是否有助于搜索获得更好的效率? |
|
7 Experimental Evaluation In all of our extended solvers, we use the following parameter values: a length of window of recent conflicts k = 50 an activity score reduction factor Q = 0.1. Source code of our CRVR extensions are available at [4]. The solver MplDL employs a combination of the decision heuristics DIST [29], VSIDS [22] and LRB [18], which are acti- Kissat-sat and Kissat-default use VSIDS and Variable Move to Front (VMTF) [26] alternately during the search. 7.1 Implementation 7.2 Experiments and Results Overall, compared to their baselines, our extensions perform better on SAT instances, but loose a small number of UNSAT instances. 总的来说,与他们的基线相比,我们的扩展在SAT实例上表现得更好,但会丢失少量的UNSAT实例。 |
|
8 Detailed Performance Analysis of CRVR |
|
9 Related Work
|
|
10 Conclusions and Future Work We present a characterization of sc and mc decisions in terms of average learned clause quality that each type produces.根据每种类型产生的平均学习子句质量,我们提出了sc和mc决策的特征。 Then we analyze how mc decisions with different bursts are distributed in CDCL search. 然后分析了不同突发的mc决策在CDCL搜索中的分布情况。 Our theoretical analysis showsthat learned clauses in a mc are connected, indicating that conflicts that occur in a mc decision are related to each other. 我们的理论分析表明,在决策过程中学习到的从句是相互关联的,这表明决策过程中发生的冲突是相互关联的。 We introduced a measure named ConflictsProximity that enables the study of proximity of conflicts in a given sequence of conflicts. Our empirical analysis shows that conflicts in mc decisions are more closely related than conflicts in sc decisions. Finally, we formulated a novel CDCL strategy CRVR that reduces the activity score of some variables that appear in the clauses learned over mc decisions. 最后,我们提出了一种新的CDCL策略CRVR,它降低了在mc决策中学习的子句中出现的一些变量的活动分数。 Our empirical evaluation with three modern CDCL SAT solvers shows the effectiveness of CRVR for the SAT instances from SAT20. In the future, we intend to pursue the following research questions:
|
|
References 1. SAT Competition 2020, http://sat2018.forsyte.tuwien.ac.at/index-2.html, accessed date: 2021-03-06. 2. SAT Competition 2020, https://satcompetition.github.io/2020/downloads.html, accessed date: 2021-03-06. |
|
文献学习——A Deep Dive into Conflict Generating Decisions的更多相关文章
- 【转载】 迁移学习(Transfer learning),多任务学习(Multitask learning)和端到端学习(End-to-end deep learning)
--------------------- 作者:bestrivern 来源:CSDN 原文:https://blog.csdn.net/bestrivern/article/details/8700 ...
- 重磅解读:K8s Cluster Autoscaler模块及对应华为云插件Deep Dive
摘要:本文将解密K8s Cluster Autoscaler模块的架构和代码的Deep Dive,及K8s Cluster Autoscaler 华为云插件. 背景信息 基于业务团队(Cloud BU ...
- 深度学习(Deep Learning)资料大全(不断更新)
Deep Learning(深度学习)学习笔记(不断更新): Deep Learning(深度学习)学习笔记之系列(一) 深度学习(Deep Learning)资料(不断更新):新增数据集,微信公众号 ...
- Deep Dive into Spark SQL’s Catalyst Optimizer(中英双语)
文章标题 Deep Dive into Spark SQL’s Catalyst Optimizer 作者介绍 Michael Armbrust, Yin Huai, Cheng Liang, Rey ...
- 学习笔记之深度学习(Deep Learning)
深度学习 - 维基百科,自由的百科全书 https://zh.wikipedia.org/wiki/%E6%B7%B1%E5%BA%A6%E5%AD%A6%E4%B9%A0 深度学习(deep lea ...
- X64 Deep Dive
zhuan http://www.codemachine.com/article_x64deepdive.html X64 Deep Dive This tutorial discusses some ...
- 转 :hlda文献学习笔记
David M.BLEI nCR文献学习笔记(基本完成了) http://yhbys.blog.sohu.com/238343705.html 题目:The Nested Chinese Resta ...
- Protocol Informatics (PI项目)【基于网络轨迹的协议逆向工程文献学习】
Protocol Informatics[基于网络轨迹的协议逆向工程文献学习]by tsy 声明: 1)本报告由博客园bitpeach撰写,版权所有,免费转载,请注明出处,并请勿作商业用途.恕作者著作 ...
- 《Docker Deep Dive》Note - Docker 引擎
<Docker Deep Dive>Note Docker 引擎 1. 概览 graph TB A(Docker client) --- B(daemon) subgraph Docker ...
- 《Docker Deep Dive》Note - 纵观 Docker
<Docker Deep Dive>Note 由于GFW的隔离,国内拉取镜像会报TLS handshake timeout的错误:需要配置 registry-mirrors 为国内源解决这 ...
随机推荐
- django限制表单上传图片的大小
django的ImageField没有提供控制上传图片的内置方法,我们可以在表单验证的过程中用clean函数来控制,搬运博客园 python小童鞋 ,可以通过重写ImageField的方法来控制上传图 ...
- Windows打开回收站的几种方式
1. 桌面双击回收站图标打开回收站,简单.快捷.方便,但在打开多个窗口时候还要最小化这些窗口以显示桌面,如果还要迅速恢复这些窗口的话也算是麻烦 2. 磁盘每个分区根目录下都有一个名称为$Recycle ...
- QMap 删除指针内容时的一个问题
在测试代码时,发现一个问题: void UserManager::removeUser(const QString &name) { QMap<QString, User*>::I ...
- Python学习:画K帮
import datetime import pandas_datareader.data as web df_stockload = web.DataReader("600797.SS&q ...
- Kubernetes学习笔记(二)
[查看pod里container的logs] kubectl logs nginx --all-containers=true #Return snapshot lo ...
- asp.net core 浏览器向服务端传递对象或对象数组参数服务端接收方式
日常开发中我们经常会在客户端向服务器端传递参数,下面以asp.net core为例 专门分享传递对象或对象数组方法 一.键值对专递对象 以下是一个表单,现在需求是将以下 表单 所有input元素以独享 ...
- viewpager加fragment可滑动加radio跟随滑动
public class MainActivity extends AppCompatActivity implements RadioGroup.OnCheckedChangeListener, V ...
- Retrofit简要分析
Retrofit是对网络请求实现了一套请求架构封装,屏蔽底层网络实现,使网络请求像调用本地接口一样 基本使用例子 public interface GitHubService {//定义request ...
- Vue 解决因网络延时 页面中的{{XXX}}模板直接展示在用户面前,配合v-cloak指令
v-cloak指令: 1.本质是一个特殊属性,Vue实例创建完毕并接管容器后,会删掉 v-cloak属性 2.使用 CSS配合v-cloak可以解决网速慢时,页面展示出{{XXX}}模板的问题 案例: ...
- Comparator与Comparable实现List中某一对象属性排序
//实体类实现 Comparable接口 进行降序排序 public class TestCompare implements Comparable<TestCompare>{ priva ...