baseline学习1

最新推荐文章于 2023-11-16 11:33:57 发布
最新推荐文章于 2023-11-16 11:33:57 发布
阅读量933 收藏
点赞数
CC 4.0 BY-SA版权
版权声明：本文为博主原创文章，遵循 CC 4.0 BY-SA 版权协议，转载请附上原文出处链接和本声明。
本文链接：https://blog.youkuaiyun.com/launch_225/article/details/12709867

1.有关的两个参数

NAME                                 TYPE        VALUE
------------------------------------ ----------- ------------------------------
optimizer_capture_sql_plan_baselines boolean     TRUE
optimizer_use_sql_plan_baselines     boolean     TRUE


SQL> create table t1 nologging as select object_id,created from dba_objects;

Table created.

SQL> alter session set nls_data_format='yyyy/mm/dd hh24:mi:ss';
alter session set nls_data_format='yyyy/mm/dd hh24:mi:ss'
*
ERROR at line 1:
ORA-00922: missing or invalid option


SQL> alter session set nls_date_format='yyyy/mm/dd hh24:mi:ss';

Session altered.

SQL> exec :object_id :=72443

PL/SQL procedure successfully completed.

SQL> exec :created :='2010/09/05 06:16:41'

PL/SQL procedure successfully completed.

SQL>
SQL> set autot traceonly exp stat
SQL> select  * from t1 where object_id=:object_id and created=:created;

no rows selected


Execution Plan
----------------------------------------------------------
Plan hash value: 3617692013

--------------------------------------------------------------------------
| Id  | Operation         | Name | Rows  | Bytes | Cost (%CPU)| Time     |
--------------------------------------------------------------------------
|   0 | SELECT STATEMENT  |      |     8 |   176 |    46   (3)| 00:00:01 |
|*  1 |  TABLE ACCESS FULL| T1   |     8 |   176 |    46   (3)| 00:00:01 |
--------------------------------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------

   1 - filter("OBJECT_ID"=TO_NUMBER(:OBJECT_ID) AND "CREATED"=:CREATED)

Note
-----
   - dynamic sampling used for this statement (level=2)
   - SQL plan baseline "SQL_PLAN_975htc2jdg7m8dbd90e8e" used for this statement


Statistics
----------------------------------------------------------
         40  recursive calls
         52  db block gets
        289  consistent gets
        181  physical reads
       5848  redo size
        406  bytes sent via SQL*Net to client
        513  bytes received via SQL*Net from client
          1  SQL*Net roundtrips to/from client
          0  sorts (memory)
          0  sorts (disk)
          0  rows processed

SQL> /

no rows selected


Execution Plan
----------------------------------------------------------
Plan hash value: 3617692013

--------------------------------------------------------------------------
| Id  | Operation         | Name | Rows  | Bytes | Cost (%CPU)| Time     |
--------------------------------------------------------------------------
|   0 | SELECT STATEMENT  |      |     8 |   176 |    46   (3)| 00:00:01 |
|*  1 |  TABLE ACCESS FULL| T1   |     8 |   176 |    46   (3)| 00:00:01 |
--------------------------------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------

   1 - filter("OBJECT_ID"=TO_NUMBER(:OBJECT_ID) AND "CREATED"=:CREATED)

Note
-----
   - dynamic sampling used for this statement (level=2)
   - SQL plan baseline "SQL_PLAN_975htc2jdg7m8dbd90e8e" used for this statement


Statistics
----------------------------------------------------------
          5  recursive calls
          0  db block gets
        257  consistent gets
         41  physical reads
          0  redo size
        406  bytes sent via SQL*Net to client
        513  bytes received via SQL*Net from client
          1  SQL*Net roundtrips to/from client
          0  sorts (memory)
          0  sorts (disk)
          0  rows processed

SQL> exec dbms_stats.gather_table_stats(user,'t1');

PL/SQL procedure successfully completed.

SQL> l
  1* select  * from t1 where object_id=:object_id and created=:created
SQL> /

no rows selected


Execution Plan
----------------------------------------------------------
Plan hash value: 3617692013

--------------------------------------------------------------------------
| Id  | Operation         | Name | Rows  | Bytes | Cost (%CPU)| Time     |
--------------------------------------------------------------------------
|   0 | SELECT STATEMENT  |      |     1 |    13 |    46   (3)| 00:00:01 |
|*  1 |  TABLE ACCESS FULL| T1   |     1 |    13 |    46   (3)| 00:00:01 |
--------------------------------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------

   1 - filter("OBJECT_ID"=TO_NUMBER(:OBJECT_ID) AND "CREATED"=:CREATED)

Note
-----
   - SQL plan baseline "SQL_PLAN_975htc2jdg7m8dbd90e8e" used for this statement


Statistics
----------------------------------------------------------
          0  recursive calls
          0  db block gets
        184  consistent gets
          0  physical reads
          0  redo size
        406  bytes sent via SQL*Net to client
        513  bytes received via SQL*Net from client
          1  SQL*Net roundtrips to/from client
          0  sorts (memory)
          0  sorts (disk)
          0  rows processed

SQL> create index idx_object_id on t1(object_id);

Index created.

SQL> create index idx_created on t1(created);

Index created.

SQL> l
  1* create index idx_created on t1(created)
SQL> select  * from t1 where object_id=:object_id and created=:created;

no rows selected


Execution Plan
----------------------------------------------------------
Plan hash value: 3617692013

--------------------------------------------------------------------------
| Id  | Operation         | Name | Rows  | Bytes | Cost (%CPU)| Time     |
--------------------------------------------------------------------------
|   0 | SELECT STATEMENT  |      |     1 |    13 |    46   (3)| 00:00:01 |
|*  1 |  TABLE ACCESS FULL| T1   |     1 |    13 |    46   (3)| 00:00:01 |
--------------------------------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------

   1 - filter("OBJECT_ID"=TO_NUMBER(:OBJECT_ID) AND "CREATED"=:CREATED)

Note
-----
   - SQL plan baseline "SQL_PLAN_975htc2jdg7m8dbd90e8e" used for this statement


Statistics
----------------------------------------------------------
         26  recursive calls
         35  db block gets
        219  consistent gets
          2  physical reads
       4056  redo size
        406  bytes sent via SQL*Net to client
        513  bytes received via SQL*Net from client
          1  SQL*Net roundtrips to/from client
          0  sorts (memory)
          0  sorts (disk)
          0  rows processed

SQL> /

no rows selected


Execution Plan
----------------------------------------------------------
Plan hash value: 3617692013

--------------------------------------------------------------------------
| Id  | Operation         | Name | Rows  | Bytes | Cost (%CPU)| Time     |
--------------------------------------------------------------------------
|   0 | SELECT STATEMENT  |      |     1 |    13 |    46   (3)| 00:00:01 |
|*  1 |  TABLE ACCESS FULL| T1   |     1 |    13 |    46   (3)| 00:00:01 |
--------------------------------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------

   1 - filter("OBJECT_ID"=TO_NUMBER(:OBJECT_ID) AND "CREATED"=:CREATED)

Note
-----
   - SQL plan baseline "SQL_PLAN_975htc2jdg7m8dbd90e8e" used for this statement


Statistics
----------------------------------------------------------
          6  recursive calls
          5  db block gets
        185  consistent gets
          0  physical reads
         96  redo size
        406  bytes sent via SQL*Net to client
        513  bytes received via SQL*Net from client
          1  SQL*Net roundtrips to/from client
          0  sorts (memory)
          0  sorts (disk)
          0  rows processed

SQL> exec dbms_stats.gather_table_stats(user,'t1',cascade=>true);

PL/SQL procedure successfully completed.

SQL> l
  1* select  * from t1 where object_id=:object_id and created=:created
SQL> /

no rows selected


Execution Plan
----------------------------------------------------------
Plan hash value: 3617692013

--------------------------------------------------------------------------
| Id  | Operation         | Name | Rows  | Bytes | Cost (%CPU)| Time     |
--------------------------------------------------------------------------
|   0 | SELECT STATEMENT  |      |     1 |    13 |    46   (3)| 00:00:01 |
|*  1 |  TABLE ACCESS FULL| T1   |     1 |    13 |    46   (3)| 00:00:01 |
--------------------------------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------

   1 - filter("OBJECT_ID"=TO_NUMBER(:OBJECT_ID) AND "CREATED"=:CREATED)

Note
-----
   - SQL plan baseline "SQL_PLAN_975htc2jdg7m8dbd90e8e" used for this statement


Statistics
----------------------------------------------------------
          0  recursive calls
          0  db block gets
        183  consistent gets
          0  physical reads
          0  redo size
        406  bytes sent via SQL*Net to client
        513  bytes received via SQL*Net from client
          1  SQL*Net roundtrips to/from client
          0  sorts (memory)
          0  sorts (disk)
          0  rows processed



--将全表扫描的PLAN_NAME=‘SQL_PLAN_975htc2jdg7m8dbd90e8e’ DISABLE

SQL> DECLARE
  2   l_plans_altered  PLS_INTEGER;
  3  BEGIN
  4   l_plans_altered := DBMS_SPM.alter_sql_plan_baseline(
  5     sql_handle      => 'SQL_93961960a2d79e68',
  6      plan_name       => 'SQL_PLAN_975htc2jdg7m8dbd90e8e',
  7     attribute_name  => 'ENABLED',
  8     attribute_value => 'NO');
  9   DBMS_OUTPUT.put_line('Plans Altered: ' || l_plans_altered);
 10  END;
 11  /
Plans Altered: 1

PL/SQL procedure successfully completed.


可以看到已经走回正常的执行计划

  1* select  * from t1 where object_id=:object_id and created=:created
SQL> /

Execution Plan
----------------------------------------------------------
Plan hash value: 2724989281

---------------------------------------------------------------------------------------------
| Id  | Operation                   | Name          | Rows  | Bytes | Cost (%CPU)| Time     |
---------------------------------------------------------------------------------------------
|   0 | SELECT STATEMENT            |               |     1 |    13 |     2   (0)| 00:00:01 |
|*  1 |  TABLE ACCESS BY INDEX ROWID| T1            |     1 |    13 |     2   (0)| 00:00:01 |
|*  2 |   INDEX RANGE SCAN          | IDX_OBJECT_ID |     1 |       |     1   (0)| 00:00:01 |
---------------------------------------------------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------

   1 - filter("CREATED"=:CREATED)
   2 - access("OBJECT_ID"=TO_NUMBER(:OBJECT_ID))




删除baseline

SQL> declare
  2  n PLS_INTEGER;
  3  begin
  4  n :=DBMS_SPM.DROP_SQL_PLAN_BASELINE (
  5  sql_handle=>'SQL_93961960a2d79e68',
  6  plan_name=> NULL)
  7  ;
  8  dbms_output.put_line(n);
  9  end;
 10  /

PL/SQL procedure successfully completed.

相对应的文档语法

DROP_SQL_PLAN_BASELINE Function
This function drops a single plan, or all plans associated with a SQL statement.
Syntax
DBMS_SPM.DROP_SQL_PLAN_BASELINE (
sql_handle IN VARCHAR2 := NULL,
plan_name IN VARCHAR2


这里有个问题，删除完baseline后，在同一个会话中，再次执行同样的主句，并不会被系统重新自动捕获；

  1* select  * from t1 where object_id=:object_id and created=:created
SQL> /

Execution Plan
----------------------------------------------------------
Plan hash value: 2724989281

---------------------------------------------------------------------------------------------
| Id  | Operation                   | Name          | Rows  | Bytes | Cost (%CPU)| Time     |
---------------------------------------------------------------------------------------------
|   0 | SELECT STATEMENT            |               |     1 |    13 |     2   (0)| 00:00:01 |
|*  1 |  TABLE ACCESS BY INDEX ROWID| T1            |     1 |    13 |     2   (0)| 00:00:01 |
|*  2 |   INDEX RANGE SCAN          | IDX_OBJECT_ID |     1 |       |     1   (0)| 00:00:01 |
---------------------------------------------------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------

   1 - filter("CREATED"=:CREATED)
   2 - access("OBJECT_ID"=TO_NUMBER(:OBJECT_ID))


而需要重新新开一个会话执行

 1* select  * from t1 where object_id=:object_id and created=:created
SQL> /

no rows selected


Execution Plan
----------------------------------------------------------
Plan hash value: 2724989281

---------------------------------------------------------------------------------------------
| Id  | Operation                   | Name          | Rows  | Bytes | Cost (%CPU)| Time     |
---------------------------------------------------------------------------------------------
|   0 | SELECT STATEMENT            |               |     1 |    13 |     2   (0)| 00:00:01 |
|*  1 |  TABLE ACCESS BY INDEX ROWID| T1            |     1 |    13 |     2   (0)| 00:00:01 |
|*  2 |   INDEX RANGE SCAN          | IDX_OBJECT_ID |     1 |       |     1   (0)| 00:00:01 |
---------------------------------------------------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------

   1 - filter("CREATED"=:CREATED)
   2 - access("OBJECT_ID"=TO_NUMBER(:OBJECT_ID))

Note
-----
   - SQL plan baseline "SQL_PLAN_975htc2jdg7m890704d19" used for this statement


Statistics
----------------------------------------------------------
          0  recursive calls
          0  db block gets
          3  consistent gets
          0  physical reads
          0  redo size
        406  bytes sent via SQL*Net to client
        513  bytes received via SQL*Net from client
          1  SQL*Net roundtrips to/from client
          0  sorts (memory)
          0  sorts (disk)
          0  rows processed



launch_225
关注
从baseline代码详解入门深度学习心得——Datawhale AI 夏令营
NoFear__77的博客
07-15 923
当前机器翻译任务的主流解决方案是。
李宏毅机器学习hw1 Boss baseline（2022）
asf2013的博客
02-26 1224
Hw1 Boss baseline（DNN二分类预测问题）
baseline学习
qq_41834327的博客
02-21 711
在学习时间序列Baseline过程中所遇到的问题有，首先是对nc格式张量的转化有了一定的了解，但是对于数据具体表示结构和索引处理方法却一知半解，比如对模型中训练集shape的转化了解不同，对于Baseline中提供的h5模型不够了解，缺乏认知，再有就是通过这次baseline的学习熟悉了天池比赛平台的任务提交整体流程，提交镜像所需要配置文件的构建流程，像result的封装方法，和压缩包的生成函数方法， test_path = './tcdata/enso_round1_test_20210201/' ##
策略梯度中的baseline
Trony的博客
09-04 1517
策略梯度中的Baseline Policy Gradient with Baseline Policy Gradient 策略梯度是关于策略网络的参数求的，策略网络π(a∣s;θ)\pi (a|s;\theta)π(a∣s;θ)的参数是θ\thetaθ，我们使用策略网络来控制Agent做运动。状态价值函数Vπ(s)V_{\pi}(s)Vπ​(s)是动作价值函数的期望，期望是关于动作A求的，动作A的概率密度函数是π\piπ，可以将期望等价写为连加的形式。这里期望中包含策略网络的参数θ\thetaθ，所以得到
idea识别不了function_视频行人重识别的一个不太强的baseline
weixin_42505252的博客
01-16 304
整合了罗浩博士的工作和高继扬博士的工作，我们现在开源了一个在目前所有的已开源可复现的视频行人重识别方法中能够达到最优的baseline，目前这个代码在MARS数据库上我本地不用rerank的最好结果是：mAP : 81.2% Rank-1 : 86.6% Rank-5 : 96.0% 在政杰同学@张政杰的实验中目前效果能达到：mAP : 81.9% Rank-1 : 87.6% yua...
baseline日常使用小结
N
10-28 3173
一. 涉及到的函数(4点)Summary of DBMS_SPM Subprograms ........................................................................................ 135-9ALTER_SQL_PLAN_BASELINE Function ............................
ffmpeg vidstabdetect的参数result指定绝对路径报错
luxianwang的博客
06-30 1740
ffmpegvidstabdetect的参数result指定绝对路径报错。 报错1：Undefined constant or missing '(' in ' XXX 解决：绝对路径冒号前加两个右斜杠，如C\\:XXX 报错2：cannot open transform file XXX 解决：（1）路径连接符使用4个或者8个右斜杆，如C\\:\\\\users\\\\XXXX （2）路径不要有中文 tips：查看源代码可知，写入trf文件是通过C语言中的fopen()...
LSTM baseline 补充学习1
08-03
当读到第 100 个单词的时候，word_100 已经记不住前面第 1个，第 2 个或者前面的单词了，因为间隔实在太远了，身处前面单词的信息早就随着不断的传递变
sqlplan baseline 学习笔记
03-02
sqlplanbaseline，绑定执行计划，固定执行计划，使用baseline固定执行计划
强化学习-学习笔记14 | 策略梯度中的 Baseline
www_xuhss_com的博客
07-12 970
本篇笔记记录学习在 策略学习 中使用 Baseline，这样可以降低方差，让收敛更快。Vπ(s)=EA∼π[Qπ(s,A)]=∑aπ(a|s;θ)⋅Qπ(a,s)V_\pi(s)=\mathbb{E}_{A\sim \pi}[Q_\pi(s,A)]=\sum\limits_{a}\pi(a|s;\theta)\cdot Q_\pi(a,s)∂ Vπ(s)∂ θ=EA∼π[∂lnπ(A|s;θ)∂θ⋅Qπ(s,A)]\frac{\partial \ V_\pi(s)}{\partial \ \theta}=\
使用GPT快速开发视频连接工具
weixin_46759000的博客
11-16 344
在网上看到了一些有趣的小视频，想着把这些视频连接成一个容易共享。可是作为一个程序员，对视频编辑软件又不了解。于是乎，我决定用程序员的方法来解决问题。考虑到这种工具编写虽然不难，但是比较烦，所以…
CUDA编译报错
Kelvin_Yan的专栏
12-02 1739
error LNK2019: 无法解析的外部符号 cudaGetDeviceProperties附加依赖项-添加 cudart.lib
深入理解PHP错误：Undefined constant
理论都是虚的，代码才是王道。
11-07 3644
其中，"Error: Undefined constant"是一个常见的错误，它表示我们在代码中使用了一个未定义的常量。当我们在PHP代码中遇到"Error: Undefined constant"错误时，我们需要首先理解这个错误的含义。然而，由于这个常量没有被声明和初始化，所以PHP会抛出一个"Error: Undefined constant"错误。现在，当我们尝试输出这个常量时，PHP就不会再抛出错误了。一旦我们发现了一个未定义的常量，我们就可以通过声明和初始化这个常量来修复这个错误。
最简单的基于FFmpeg的AVUtil例子 (AVLog, AVOption等)
雷霄骅(leixiaohua1020)的专栏
07-18 1万+
本文的示例程序记录了FFmpeg的libavutil中几种工具函数的使用方法：AVLog：日志输出AVOption (AVClass)：选项设置AVDictionary：键值对存储ParseUtil：字符串解析几个libavutil的工具AVLogAVLog是FFmpeg的日志输出工具。在FFmpeg中所有的日志输出不是通过printf()函数而是通过av_log()函数。av_log()会最终调
CheckStyle报错的常见问题及解决方式
热门推荐
10-19 5万+
CheckStyle报错的常见问题及解决方式
vue全家桶使用遇到的eslint问题以及解决方案
Stars_in_rain的博客
07-15 1万+
1.Trailing spaces not allowed 去掉多余的空格 2. Newline required at end of file but not found 需要在最后的后面回车，空一行 3.Missing space before value for key ‘name’ 在关键字“值”之前缺少空格 4.space-before-blocks 关键字后面要空一格。 5.A space is required after ‘,’ 在逗号后面要加空格 6.no-unused-vars 不能有声
避免出现"Notice: Use of undefined constant......"警告
hairi的专栏
03-23 2万+
刚开始学习php，写了一段测试程序，如下：$username = $_POST[username];//从html表单获取数据；$password = $_POST[password];$link = mysql_connect(localhost,root,123) or die(could not connect:.mysql_error()) ;$select = m
深度学习 baseline
最新发布
03-29
### 深度学习中的 Baseline 模型 在深度学习的研究和应用过程中，Baseline 是指一种用于衡量其他更复杂模型性能的基础模型或方法。它通常是一个简单的、易于实现的模型，能够提供一个初步的结果作为后续优化的目标参考点[^2]。 #### 定义与作用 Baseline 的主要目的是为新模型的设计和改进提供一个清晰的标准。通过对比不同模型的表现，研究人员可以判断新的架构或算法是否有显著提升。例如，在图像分类任务中，可以选择逻辑回归或者卷积神经网络（CNN）作为一个基础模型，并将其结果视为基准线。随后在此基础上尝试引入更深的网络结构或其他技术手段来进一步改善效果[^3]。 #### 构建方式 构建 Baseline 可以采用多种策略： - **预训练模型**：如果存在针对特定领域已有的高质量预训练权重，则可以直接加载这些参数并微调至当前任务上； - **简单模型**：对于某些场景来说，即使是最基本的传统机器学习算法也可能成为有效的起点，比如支持向量机(SVM) 或随机森林(Random Forests)； 当涉及到具体应用场景时，还需要考虑数据特性以及计算资源等因素综合决定最适合自己的方案。 ```python from sklearn.linear_model import LogisticRegression from keras.models import Sequential from keras.layers import Dense, Conv2D, Flatten # Example of a simple logistic regression as baseline model def create_logistic_baseline(): lr = LogisticRegression() return lr # Example of CNN-based deep learning baseline using Keras framework def build_cnn_baseline(input_shape): model = Sequential() model.add(Conv2D(32,(3,3),activation='relu',input_shape=input_shape)) model.add(Flatten()) model.add(Dense(units=10, activation="softmax")) return model ``` 上述代码展示了如何创建两个不同的基线模型——一个是传统的逻辑回归，另一个则是基于卷积层的小型神经网络。这两种都可以视作各自类别下的初始参照物。
评论
添加红包

请填写红包祝福语或标题

个

红包个数最小为10个

元

红包金额最低5元

当前余额3.43元 前往充值 >
需支付：10.00元
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付元
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明：

1.余额是钱包充值的虚拟货币，按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载，可以购买VIP、付费专栏及课程。

余额充值