- 博客(96)
- 收藏
- 关注
RSS订阅原创 Install cuDNN v8.0.5 in ubuntu 20.04
what: NVIDIA cuDNN是用于深度神经网络的GPU加速库。cuDNN安装很简单，就是将解压的文件放到指定位置即可.why: GPU acceleration for DNNhow: cuDNN安装很简单，就是将解压的文件放到指定位置即可[2].step 1: Register a Nvidia account and go to the download page. Select the version based on your CUDA versionstep 2: naviga.
2021-01-24 00:01:27
819
原创 [210121] Nvidia驱动自动更新后导致nvidia-smi不可用
[210121] Failed to initialize NVML: Driver/library version mismatchwhat: failed to use nvidia-smi on 1/21, but it worked normally on 1/20.(base) shane@shane-MS-7A94:~$ nvidia-smiFailed to initialize NVML: Driver/library version mismatchwhy: Driver/li
2021-01-22 11:03:54
623
原创 [210121] install nvidia driver via ssh
[210121] install nvidia driver via sshwhat: install nvidia driver via sshwhy: server is in another placehow:step 1: download target driverstep 2: stop X server service# 为了能够安装新的 nvidia 驱动，我们需要关闭这个 X 服务，我的 Ubuntu 是基于 lightdm 类型的桌面系统，采用如下命令：$ sudo /e
2021-01-22 10:49:10
394
1
原创 WWH3R: A Six-Step Standard Operating Procedure for Doing Things
In this report, we present a standard operating procedure (SOP) called WWH3R for doing things,which consists of six steps: What, Why, How, Record, Result, Reference.This SOP can be applied to many fields such as running projects, resolving issues,making
2021-01-19 11:00:19
157
原创 Ross Girshick: writing good research papers
[210112] Ross Girshick: writing good research paperswebsite: https://zhuanlan.zhihu.com/p/104279421What did I learn from your paper?A paper should be about a single focused idea or question.“Idea” usually means method. What should I learn?Under what
2021-01-12 15:48:40
102
原创 ISA,CISC, RISC, x86, ARM
文章目录[201214] What is ISA? Why it is so important for IC design?ISA与微架构ISA是一种接口[201214] CISC vs. RISCISA的分类[201214] x86 vs. ARM architecture?[201214] What is ISA? Why it is so important for IC design?The main ISAs used today are x86, ARM, ARC, MIPS and Po
2020-12-14 15:50:43
963
原创 “_“, “__“ and “__xxx__“ in Python?
ItemUsageDescriptionNotes_attribute保护变量类实例和子类实例能访问到这些变量需通过类提供的接口进行访问；不能用’from module import *'导入__attribute私有变量类对象自己能访问__func__系统定义代表python里特殊方法专用的标识如 __init__（）代表类的构造函数默认情况下，Python中的成员函数和成员变量都是公开的(public),在python中没有类public,p...
2020-12-08 16:12:12
62
原创 how to understand “super(classname, self).__init__()“ in Python?
# python中的super(Net, self).__init__()# 首先找到Net的父类（比如是类NNet），# 然后把类Net的对象self转换为类NNet的对象，# 然后“被转换”的类NNet对象调用自己的init函数, # see <https://blog.csdn.net/wltsysterm/article/details/104440387>super(classname, self).__init__() # required in Python2In P
2020-12-08 16:04:49
88
原创 Python: *args, **kwargs, **vars(args)?
variablesfunctionnotes*args*args(1) 表示任何多个无名参数，它本质是一个 tuple告诉函数，按tuple来解析args**kwargs**kwargs(1) 表示任何多个关键字参数，它本质上是一个 dict告诉函数，按dict来解析args，如果同时使用 *args 和 **kwargs 时，必须 *args 参数列要在 **kwargs 之前**vars(args)这与 **kwargs是类似的，告诉函数按dict来解析，还多...
2020-12-08 13:23:43
2257
原创 understand “*“ and “**“ in Python
No.categorydescriptionnotes1operator* is multiply, ** is square算数运算, *代表乘法, **代表乘方2formal argumentunknown numbers of arguments, * for decoding as tuple, ** for decoding as dict函数形参，*args 和 **kwargs 主要用于函数定义, 实现将不定数量的参数传递给一个函数3actua...
2020-12-08 12:47:09
78
原创 Python if __name__ == ‘__main__‘
(1) “__name__” 是内置变量，用于表示当前模块的名字，同时还能反映一个包的结构。(2)如果一个模块被直接运行，则其没有包结构，其 “__name__ ”值为“ __main__”.理解1：简单理解如果模块是被直接运行的，则代码块被运行，如果模块是被导入（import）的，则代码块不被运行。理解2：模拟的程序入口对于很多编程语言来说，程序都必须要有一个入口，比如 C，C++，都需要有一个main 函数来作为程序的入口。Python 则不同，它属于脚本语言，不像编译型语言那样先将程
2020-12-08 10:08:12
99
原创 PlantUML in 3 steps (VSCode)
After install Extension of PlantUML in VSCode, follow these 3 steps to export your first PlantUML picture.Step1: create file[1]代码的文件的扩展名可以如一下：*.wsd, *.pu, *.puml, *.plantuml, *.iuml.Step2: a simple example[1]@startumlRed -[#red]> Green : To startG
2020-11-27 09:55:07
136
原创 [koroFileHeader] Create my own file header in VSCode
[201124] Create my own file header in VSCodeStep1: Install extension of koroFileHeader, and edit the following default json to target.Step2: Go to setting page by typing file header in “Search settings” column, and choose Fileheader: Custom MadeStep3:
2020-11-24 17:51:18
67
原创 [201117] 为什么range范围是左闭右开区间？
[201117] 为什么range范围是左闭右开区间？range(i,j) 相当于 [i,j)这实际上遵守的是不对称边界的设计思想，是一种编程技巧，其根本原因是为了解决程序设计中难于察觉的“栏杆错误”，也叫“差一错误”（off-by-one error）(参见《C陷阱与缺陷》，p46)。Python的数组下标是从0开始的，这一点和C语言是相同的。不对称边界的编程技巧，有两种表示方式：用第一个入界点和第一个出界点来表示一个数值范围。用第一个被占用的元素和第一个被释放的元素分别来表示上界和下界。
2020-11-17 14:41:49
1560
原创 [201116] What is meta learning?
Meta learning is Learn to learn based on tasks (machine learning is learn to do a given task based on dataset).It does mean that it can handle new tasks with existing status, but it is likely to learn faster and better,because it is a good learner after
2020-11-17 06:27:32
112
原创 Transfer learning (explore possible source data to improve performance under given target data)
Use relatively abundant source data to train better model considering limited target data.Firstly, use source data to train a initial model.Secondly, use target data to re-train model (Should it be similar to fine-tune?)One case of transfer learning is
2020-11-17 06:20:41
71
原创 [201106] LaTeX use note
[201028] Table或者Figure后面加上[htb]?其中h表示here, t –top, b-bottom，即表格在文中的位置。(https://blog.csdn.net/kmsj0x00/article/details/82772000)[201028] 表格的对齐与内边框{tabular}后面的{|l||r|r|r|c|}表示对齐与边框。l-left, r-right, c-center.有’|'表示有边框，没有则表示无边框。(https://blog.csdn.net/kmsj
2020-11-06 13:54:49
134
原创 Summary on VoxSRC20
My simple storyOne laymen and hobbyist, start from zeroLearn and use deep learningJoin the first challenge on speaker recognition as practiceStatus/TaskStatus: Use part-time to learn and practice.Task: Join VoxSRC20 and make it to TOP10ActionU
2020-11-04 12:09:01
101
原创 Error - use ‘,‘ as separator of authors in making reference.bib
Waste at least 2 hours to make reference.bib due to this error.The correct separator of authors is ‘and’.Lesson learndon’t use ‘,’ as separator of authors, you should use and
2020-10-30 13:37:33
45
原创 LaTeX Error: Something‘s wrong--perhaps a missing \item
Got this error when making ‘reference.bib’.Analysis: typed the error message into Bing, and got similar reported issuesResolution: you need at least one \cite (or \nocite) command in your document
2020-10-30 13:35:50
204
原创 Recipe terminated with fatal error: spawn latexmk ENOENT.
AnalysisVersion info: VS code:1.50.0, Tex Live:2019/Debian, LaTex Workshop:8.13.2ActionMake sure your operation is correct by referencing to official guideline and VS Code + LaTeX.Install TexLive: TeXLive即LaTeX的编译环境, Use latex --help or latex -versio.
2020-10-27 07:18:09
12773
原创 正斜杠与反斜杠
Q: What is the difference between Forward slash ‘/’ and backward slash ‘’?A: Answer in (https://www.cnblogs.com/xyao1/p/10916090.html)\n, 除了在程序中，其它地方几乎不需要用\，应该就差不多了，\是转义字符除法与分隔符, 在绝大多数地方，用的都是/（slash），包括Mac/Linux，也包括URL。Windows崛起你唯一需要记住的是，Microsoft这个怪鸡
2020-10-26 16:22:54
151
原创 How to embed used fonts while making PDF file in Linux?
Take Ubuntu 20.04 for example:$ pdftops intput_file.pdf$ ps2pdfwr -dPDFSETTINGS=/prepress intput_file.ps output_file.pdfI tried, and got succeeded.
2020-10-25 08:28:55
51
原创 General architecture of technical report
TitleAbstract1. IntroductionProblem statementResolution of this paperContribution2. Related work / Prior work (optional)3. Proposed method / Approach4. ExperimentsDatasetsTrainingResults5. Future work / Discussion (optional)6. ConclusionA
2020-10-22 11:33:48
53
原创 [201021] difference between maximum likelihood and minimum cross entropy?
[201021] difference between maximum likelihood and minimum cross entropy?A: no difference, they are the same during training neural network.
2020-10-21 10:41:04
49
原创 [201021] how to judge over-fitting or under-fitting?
[201021] how to judge over-fitting or under-fitting?A: See the performance on training set and validation set.train setdev setstatusgoodbadover-fittingbad–under-fittingFor under-fitting case, the model may be too poor.
2020-10-21 10:39:15
67
原创 Better understanding of EFR and EFA of binary classifier
Better understanding of EFR(错误拒绝率 ) and EFA(错误接受率)[1][1] 说话人确认系统性能评价指标
2020-10-15 14:22:31
60
原创 [200919] My way to learn new thing - FPL-TARGE
[200919] My way to learn new thing - FPL-TARGEFPL-TARGE: Focus, Learn, Practice, Think, Awareness, Record, Go over, Endure:Focus: direction is always the most important thing, once determined, never hesitatePractice: Practice is more important than lea
2020-10-15 12:35:52
83
原创 [200917] difference between ‘/‘, ‘./‘ and ‘../‘
[200917] difference between ‘/’, ‘./’ and ‘…/’‘/’ is the root folder‘./’ is current folderfor example: python test.py and python ./test.py is the same‘…/’ is parent folder
2020-09-17 10:42:21
62
原创 [voxsrc20_ask_27] How to get DET plot for a speaker recognition system?
ID = voxsrc20_ask_27Status: closedQuestionHow to get DET plot for a speaker recognition system?DET plot: Detection Error Trade-off plot[1]AnswerUse python to make it, see [2].Notes: Tested okay.import matplotlib.pyplot as pltfrom sklearn.metrics
2020-09-16 12:44:07
170
原创 [python_tsk_00] Install numpy and scipy in Win10 via pip3
ID = python_tsk_00TaskStatus: Got this error ModuleNotFoundError: No module named 'numpy'Task: To install numpy and scipy in Win10.ActionUse pip3 to make it, and referenced to this linkThis is the step-by-step operation record:Step1: Install and che
2020-09-15 16:45:13
127
原创 [200908] 熵，交叉熵，KL散度
[200908] 熵，交叉熵，KL散度信息熵（Entropy）信息熵（Entropy）是衡量分布的混乱程度或分散程度的一种度量。分布越分散(或者说分布越平均)，信息熵就越大。分布越有序（或者说分布越集中），信息熵就越小。(https://www.cnblogs.com/soyo/p/6893551.html)信息熵用bit位数来定量表示。H(p)=−∑pi∗log2pi H(p) = -\sum{p_i*\log_2{p_i}}H(p)=−∑pi∗log2pi其中pip_ipi为X中
2020-09-08 15:02:47
196
原创 [200903] 逻辑回归及交叉熵
[200903] 逻辑回归及交叉熵通过B站，理解了逻辑回归和交叉熵：Q: 逻辑回归为什么不可以叫逻辑分类？A: 用回归的方法（找到不同类别之间的分界线）解决逻辑分类问题（包括二分类和多分类问题）。基于sigmoid的二分类问题的交叉熵sigmoid的输出是[0,1]这是一个概率的范围。对于二分类问题，给定一个数据[x,ylabely_{label}ylabel], 预测值为 ypredy_{pred}ypred.当ylabely_{label}ylabel为0时，损失函数表示为-log(
2020-09-03 13:33:30
113
原创 [200902] 全概率与贝叶斯公式
[200902] 全概率与贝叶斯公式通过B站，巩固了全概率公式与贝叶斯公式：首先是某个事件B的完备事件组{A1,A2,…,An}可以用因果关系来理解。Ai是第i个“因”，B是“果”。Ai --------> B 代表Ai发生后B发生的概率，即条件概率P(B|Ai)B ---------> Ai 代表B发生了，同时是由Ai发生导致的概率，即P(Ai|B)全概率公式先验概率范畴：由因及果。在Ai发生的条件下，所有Ai–>B的可能性，加在一起。P(B)=∑P(B∣Ai)∗P(
2020-09-03 13:13:12
256
原创 [200902] 条件概率与独立事件
[200902] 条件概率与独立事件通过B站，巩固了条件概率与独立事件：条件概率表示事件A发生后，事件B发生的概率。两者应该处于同一空间。A与B同时发生的概率，除以A发生的概率。P(B∣A)=P(AB)P(A) P(B|A) = \frac{P(AB)}{P(A)} P(B∣A)=P(A)P(AB)A和B的空间来理解，如果A和B是互斥事件，则P(AB)=0。如果两者独立，则P(B|A)=P(B)，即A发生与否对B无影响。如果两者有交集，则条件概率可以表征两者交集的大小。如果条件概率越大，则
2020-09-03 11:15:32
2281
原创 [200902] 协方差与相关系数
[200902] 协方差与相关系数通过B站，学习到协方差与相关系数。有如下几个要点：协方差是两个变量之间的总体误差，方差是协方差的特例，即两个变量相同时。Cov(X,Y)=E(X−E(X))∗E(Y−E(Y)) Cov(X,Y)=E(X-E(X))*E(Y-E(Y)) Cov(X,Y)=E(X−E(X))∗E(Y−E(Y))特别的：Cov(X,C)=0, C为常数;Cov(aX,bY)=ab*Cov(X,Y)相关系数是表示两个变量之间的线性相关度的，范围为[-1, 1]。0为不相关，正数为正相
2020-09-03 11:14:07
153
原创 [200830] Temperature monitoring in Linux
[200830] temperature monitoringCPUUse ‘lm_sensors’ to monitor CPU: Linux 查看CPU温度# checkrpm -q lm_sensors# installsudo apt-get install lm-sensors -y# configure sensorsyes | sudo sensors-detect# runsensors# watch every 2swatch -n 2 sensorsGPUw
2020-08-30 13:39:30
103
原创 voxsrc20_tsk_03-用ZeroTier搭建虚拟局域网实现内网穿透
ID = voxsrc20_tsk_03Status: closed文章目录精华TaskExecution[200816] First try1. Use requirement2. Solution3. How to?Step1: Register account of zerotier and create your NetworkStep2: install zerotier in serverStep3: install zerotier in notebookStep4: join to ta
2020-08-25 14:46:14
471
原创 [200823] how to understand the impact of bias and variance?
[200823] how to understand the impact of bias and variance?The complexity of model is depended on the dataset size.dataset sizemodel complexitybias/varianceresultsmallsimplebalancedgoodsmallcomplexvarianceover-fitting (good @training
2020-08-23 22:33:42
69
原创 [200814] 自己动手，搞定软件著作权申请（已成功）
写在前面：为什么我要自己搞软件著作权？因为有需要，所以要搞x宝上找一找，30天可以搞定，只要收费400。何苦自己弄呢？一个字：穷 + 图个心安精华如下：单人申请材料申请表（单面打印，第3页签名，第4页手写材料页数）源码文档（单面打印）软件设计文档或者使用说明文档（单面打印）个人身份证复印件（正反面都要复印）文章目录Step1: 官网注册Step2: 填写申请表Step3: 准备说明书Step4: 准备源程序Step5: 检查常见错误Step6: 各种情况所需材料清单列表（
2020-08-14 14:35:10
1153
空空如也
空空如也
TA创建的收藏夹 TA关注的收藏夹
TA关注的人