- 博客(37)
- 收藏
- 关注
RSS订阅原创 Principal Components Analysis: PCA 主成分分析
math basicsorthornormal basis identity matrix (a special othornormal basis)variance (square of standard deviation) and its goal 1. noise: noise should contains low or no information needed to be extracted, it can be quantified by SNR....
2020-07-21 08:28:12
483
1
原创 FPGA in deep learning: computer vision application
note: I am only interested in FPGA application in inference. I am still believing GPU a better candidate for training.reference : https://software.intel.com/en-us/ai/courses/deep-learning-inferen...
2019-05-28 07:23:47
219
原创 FPGA in deep learning: introduction
note: I am only interested in FPGA application in inference. I am still believing GPU a better candidate for training.reference : https://software.intel.com/en-us/ai/courses/deep-learning-inferen...
2019-05-28 06:55:30
295
原创 deep learning: deep feedforward network (output and hidden layer)
deep forward networks = feedforward neural networks = multiplayer perceptrons (MLP)input layer, hidden layer, output layer; activation function; cost function...goal: define a mapping and learns ...
2019-02-04 08:51:22
772
原创 MLaPP: 02 Probability (joint probability)
covariance and correlation:for two univariate X and Y, the covariance is defined as:if X and Y are independent, then covariance is zero and correlation is also zero; but the converse is not ...
2018-12-30 16:26:07
379
原创 MLaPP: 02 Probability (univariate)
probability interpretation:frequentist interpretation: probabilities represent long run frequencies of events. for example: if we flip the even coin many times, we expect it to land heads about 1/2 ...
2018-12-30 14:37:15
257
原创 MLaPP: 01 Introduction
types of machine learning:predictive or supervised learning descriptive or unsupervisoed learning reinforcement learningsupervised learning:classification:inputs: xoutputs: ywhere , with ...
2018-12-30 12:23:56
605
原创 线性代数基础(3)
一个矩阵A可逆当且仅当它的行列式detA非零。若A为三角阵，则detA等于A的主对角线上元素的乘积。， adjA是一个伴随矩阵。
2018-12-28 03:46:26
809
原创 线性代数基础(2)
分块矩阵：对于稀疏矩阵，上面有许多的0，将变量正确分组可以产生许多的0方块的分块矩阵。矩阵分解：为进一步简化计算，可以对矩阵进行LU分解。可逆矩阵：可逆矩阵定理只能用于方阵。分块矩阵：分块矩阵可以进行乘法，只要A的列的分法与B的行的分法一致。矩阵因式分解：LU分解：分解算法： R...
2018-12-28 03:33:29
263
原创 线性代数基础(1)
线性方程：矩阵方程 Ax=b齐次线性方程组：非齐次方程组：线性无关：线性变换：通常把矩阵A当做一种“对象”，它通过乘法“作用”于向量x，产生的新向量叫做Ax.: ...
2018-12-27 16:19:58
139
原创 PRML-3
given a trainning data set comprising N observations {Xn} together with corresponding target values {tn}, the goal is to predict the value of t for a new value of x. from a probabilistic perspective, ...
2018-12-27 08:20:44
243
原创 线性代数基础(6)
内积（点积）：向量的长度：Rn空间中的距离：正交向量：正交集：正交投影：当y表示成Rn空间中u1,u2,...un的线性组合时，y中的各项可以分成两部分，使得y可以写成：y=z1+z2。其中z1是ui中的某些线性组合，z2是其余的ui的线性组合。当{u1,u2...un}是正交基时，这个思路特别有用。...
2018-12-27 07:49:36
197
原创 线性代数基础(5)
特征值：定理：三角矩阵的主对角线的元素是其特征值以下数值方程称为矩阵A的特征方程：对角化：矩阵的对角化可以分四步：1）求出A的特征值2）求A得线性无关的特征向量3）利用2）中的特征向量构成矩阵P4）用对应的特征值来构造矩阵D（对角矩阵）对于有n个相异的特征值的n x n的矩阵是可以对角化的。特征值得迭代估计：幂...
2018-12-25 05:49:12
165
原创 线性代数基础(4)
向量空间定义：向量空间是由一些被称为向量的对象构成的非空集合V。定理：零空间，列空间，线性变换一般的，子空间是通过常用的两种方式产生：1）齐次线性方程组的解集；2）某些确定向量的线性组合的集合。齐次线性方程组Ax=0的解集x可以认为是矩阵A的零空间。对于m x n矩阵A的零空间可以写成NulA := {x: Ax=0}. 可以更进一步的描述为Rn中在线性变换x-->Ax下映...
2018-12-24 14:47:33
2467
原创 python: understand super()
super() function allows the subclass to call the superclass methods explicitly, useful in inheritance. note: there is some difference between python2 and python3 in use of super().here is an examp...
2018-12-12 15:44:16
268
原创 python: understand 'self' and '__init__'
reference:https://micropyramid.com/blog/understand-self-and-__init__-method-in-python-class/ https://stackoverflow.com/questions/8609153/why-do-we-use-init-in-python-classes Before understanding...
2018-12-10 16:49:35
195
原创 PRML: ML 估计 (1)：简介
前提条件：i.i.d 模型已知，参数未估基本思想：采样样本得到数据集，选取合适的模型，采用最大似然的思想估计参数值。以高斯分布为例:采样数据集的概率分布是i.i.d. 从而有：利用最大似然估计的思想：将似然性最大化 -->对未知参数求导得到概率分布的全局(如果有)最大值 -->对下面分布函数分别对期望和方差求导：从而得到：注意：上述的期望是...
2018-12-10 02:26:57
2789
原创 FPGA clock network architecture: Flexibility vs. Area and Power (paper)
commercial clock networks:current commercial FPGA support multiple and local and global clock domains. the FPGA is divided into four quadrants.Altera Straitix II provides 16 global clock signals, ...
2018-12-08 14:15:15
299
原创 FPGA architecture overview
programming technology:static memory (RAM): main stream methodology. the SRAM used for: programming the routing interconnects that are usually steered by small multiplexors programming CLBs that...
2018-12-03 15:28:35
998
转载 FPGA clock resources
good article cited from here: https://www.allaboutcircuits.com/technical-articles/clock-management-clock-resources-of-fpgas/This article will review the common FPGA resources that allow us to effici...
2018-12-02 06:49:54
885
原创 python: data structure - Queue
if dealing with millions data, an effcient container is needed: queue List: slow for inserting; Set: slow for checking, extension, searching, and soring Queue: FIFO and LIFOFIFO queues (queue)use...
2018-11-25 08:33:10
225
原创 python: data structure-list
List is a collection which is ordered and changeable. Allows duplicate members. Tuple is a collection which is ordered and unchangeable. Allows duplicate members. Set is a collection which is uno...
2018-11-21 13:36:42
208
原创 python: data structure - dictionaries
dictionaries are extremely efficient at looking up a value, given a specific key object that maps to that value. there is a key and the corresponding value associated with that key for each dictionary...
2018-11-20 14:23:46
346
原创 python: data structure-tuples and namedtuples
tuplestuples are objects that can store a specific number of other objects in order, and are immutable. the benefit of tuple's immutability is they can be used as keys in dictionaries, and in other ...
2018-11-20 13:31:15
171
原创 python: __name__
simply say:the __name__ always points to the name space of the currently running module. so, if it's running in an imported module, then calling __name__ would return the name of that module; if it'...
2018-11-19 02:48:17
160
转载 Python 3's f-Strings: An Improved String Formatting Syntax (Guide)
source: https://realpython.com/python-f-strings/As of Python 3.6, f-strings are a great new way to format strings. Not only are they more readable, more concise, and less prone to error than other w...
2018-11-19 02:36:37
410
转载 python: @property
source: http://python.jobbole.com/80955/ and http://www.blog.pythonlibrary.org/2014/01/20/python-201-properties/Python has a neat little concept called a property that can do several useful things. ...
2018-11-18 03:09:41
114
原创 python: hasattr()
The hasattr() method returns true if an object has the given named attribute and false if it does not.# hasattr(object,name)class FPGA: dev_type = "arria" dev_pack = "fbga"a...
2018-11-14 12:27:41
474
转载 The Four Stages of Compiling a C Program
https://www.calleerlandsson.com/posts/the-four-stages-of-compiling-a-c-program/Knowing how compilation works can be very helpful both when writing code and when debugging.Compiling a C program is ...
2018-09-15 12:23:59
188
原创 pyopencl 常用的id
source: http://blog.sina.com.cn/s/blog_539a3a3201011ld9.htmlsource: https://jorudolph.wordpress.com/2012/02/03/opencl-work-item-ids-globalgrouplocal/global ID: image all the work items (AMD) or thread...
2018-05-29 06:02:43
468
原创 pyopencl 常用的syntax
import numpy as npa=np.random.rand(m*n).astype(np.float32)# create a m*n array, initialized with random float32 precision dataimport pyopencl as clctx=cl.create_some_context() #create space for the...
2018-05-29 04:44:33
576
原创 pyopencl architecture
header:import pyopencl as clchoose the platform that support the OpenCL you want, for example, platform '0' here in default (since I have only one laptop connected):platforms=cl.get_platforms()[0]it w...
2018-05-29 01:39:46
420
原创 image processing: restoration
restoration in the presence of noise only-spatial filteringMean filter: all kinds of mean filters. overall the geometric mean filters are better suited for random noise like Gaussian or uniform noise....
2018-05-28 01:32:51
384
原创 computer vision: background
1. computer vision: trying to inverse what we see in one or more images and to reconstruct its properties, such as shape, illumination, and color distributions.2. computer vision applications:optical ...
2018-05-27 09:48:23
264
原创 algorithm: face detection
coupled with color-based clothing and hair detection algorithms, can locate and recognize the individuals.
2018-05-27 09:18:13
254
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