xiaodouhao123456的博客

神经网络的前馈过程可以得到输入向量的反馈结果。那么为了更好的训练神经网络，我么需要定义误差函数，对于单层感知机，我们定义的误差函数如下：对于多层感知机，我们定义的误差函数如下：上面的多层感知机就分为了三个层：输入层，隐藏层，输出层下面我们就学习多层的数学及代码实现。...

但是现实世界中有很多不能用直线来分类的例子，我们需要用更高维来表达。即区域。so we’re ready to put these building blocks together, and build great Neural Networks! (Or Multi-Layer Perceptrons, however you prefer to call them.)将线性模型转换为非线...

And it basically goes like this:Take your dataPick a random modelCalculate the errorMinimize the error, and obtain a better modelEnjoy!Calculating the Error Function我们看到误差函数是交叉熵的平均值。y^\hat{y...

最大似然函数用概率来表示每个点是蓝色点还是红色点的可能性，如下图所示：那么对应他们真正的颜色，四个点是他们真实颜色的概率分别如下：那么根据上面的概率分配方法，这四个点均能被正确分配颜色的概率是：那么对于另一种方法，分配的颜色接近真实的概率是：所以右侧的模型更能正确分配点，如果我们从左侧的模型开始，一步步改变模型，使其能够变成右侧的模型，这一过程称为最大化似然函数上述的例子中只有四个点，...

上一节中，我们对输出类型只有0/1两类的分类器进行了建模，那么对于多种类型的分类呢，这节将进行讲解：多分类的归一概率表示如果我们有下面三种动物，各自的分数如下，如何将分数转变为和为1的概率小数呢？上面就是归一函数的雏形，那归一函数的一般表达式如下：用代码来表达归一函数如下：def softmax(L): expL = np.exp(L) sumExpL = sum(exp...

事实上，如果我们对上一篇文章提到的点的分类问题建立误差评判函数的话，如果我们用错误点个数来作为评判标准的话，那么线在一个小的step下是无法改变错误点数的，所以算法有可能会让线左右摇摆始终找不到合适的方向去移动，所以误差评判函数必须是连续的可微的，所以我们用所有点距离线的错误距离来评判在有点未被分类正确的时候，错误值会很大，为了减少这个错误值，算法去寻找能够降低错误值的移动方向，按照降低梯度最大...

212113

You’ll have the chance to create the perceptrons for the most common of these, the AND, OR, and NOT operators. And then, we’ll see what to do about the elusive XOR operator. Let’s dive in!AND Percept...

Linear Regression（线性回归）Linear regression helps predict values on a continuous spectrum, like predicting what the price of a house will be.那么，关于将数据在离散类中归类呢？Logistic Regression (逻辑回归)it’s a classifi...

two steps:you drive and record training laps in simulationbuild and train a deep neural network that learns from the way you drive深度学习是使用深度神经网络进行机器学习的一种方法，通过深度神经网络 深度学习可以完成从语音识别到汽车驾驶等大量的任务。首先学习感...

I write Python project in this “Advanced Lane Finding” , and set the program structure like this:including the main.py and sub_functions in the module_py.now I will introduce the logic in the above...

Tips and Tricks for the ProjectCamera CalibrationYou need to set your chessboard size to 9x6 for the project instead of 8x6 as in the lesson.Do your curvature values make sense?a good check on whe...

In the next frame of video you don’t need to do a blind search again, but instead you can just search in a margin around the previous lane line position,like in the above image. The green shaded area ...

review steps and prepare for the projectSteps we’ve covered so far:Camera calibrationDistortion correctionColor/gradient thresholdPerspective transformAfter doing these steps, you’ll be given tw...

combined color and gradient thresholds for lane detectionYou can clearly see which parts of the lane lines were detected by the gradient threshold and which parts were detected by the color threshold...

不同的色彩空间A color space is a specific organization of colors; color spaces provide a way to categorize colors and represent them in digital images.RGB:red,green,blueFor example, white has the coordin...

goalNow consider how you can use various aspects of your gradient measurements (x, y, magnitude, direction) to isolate lane-line pixels.

principleIn the case of lane lines, we’re interested only in edges of a particular orientation. So now we will explore the direction, or orientation, of the gradient.functionnp.arctan2（）which ca...

goal在上一个练习中，我们看到x或者y方向的梯度会把一些图像清除掉，所以在这个练习中，我们apply a threshold to the overall magnitude of the gradient, in both x and y.（求x y的整体渐变幅度）principleThe magnitude, or absolute value, of the gradient is...

利用梯度来检测斜率较大的边缘，他们更可能是车道线。

计算道路曲率的steps知道汽车的速度，行驶状态以及车道的弯曲程度就可以计算出这个角度。所以我们需要确定道路的弯曲程度，即曲率。利用掩膜和阈值化技术来检测车道线进行透视变换获得车道的鸟瞰图对车道线进行多项式拟合，从多项式中可以获得车道线的曲率。For a lane line that is close to vertical, you can fit a line using thi...

three step cycle for roboticssensor and perceive the worlddecide what to do based on that perceptionperform an action to carry out that decision80% of building a self-driving car is perception, ...

toolsgithubIf you are unfamiliar with GitHub , Udacity has a brief GitHub tutorial to get you started. Udacity also provides a more detailed free course on git and GitHub.https://blog.udacity.com/...

Throughout this Nanodegree Program, we will be using Python with OpenCV for computer vision work. OpenCV stands for Open-Source Computer Vision.主页：https://opencv.org/canny算法边缘检测：识别图像中物体的边界，是opencv中...

identifying lane lines by colorone image including red green blueeach of these color channels contains pixels whose values range from zero to 2550 is the darkest value255 is the brightest calue

approachesself_driving car has two approaches now:robotics approach——fuses output from a suite of sensors to directly measure the vehicle surroudings and then navigate accordingly.deep learning ap...