andrew ng stanford machine learning

It is decided that machine learning is under system learning, and Stanford courseware is the main line. Notes1 is part of the http://www.stanford.edu/class/cs229/notes/cs229-notes1.pdf about Regression 1. Linear Regression For example, if the House Price is predicted and the data cannot be found on the Internet, use

invoking the example in MATLAB above, we can define the cost function of the logistic regression as follows:In the figure, Jval represents the cost function expression, where the last item is the penalty for the parameter θ; The following is a gradient of the derivation of each θj, where θ0 is not in the penalty, so gradient is not changed, and Θ1~θn has one more (λ/m) *θj respectively;At this point, regularization can solve the linear and logistic overfitting regression problem ~

two classification problem, so the model is modeled as Bernoulli distributionIn the case of a given Y, naive Bayes assumes that each word appears to be independent of each other, and that each word appears to be a two classification problem, that is, it is also modeled as a Bernoulli distribution.In the GDA model, it is assumed that we are still dealing with a two classification problem, and that the models are still modeled as Bernoulli distributions.In the case of a given y, the value of x is

calculates the accuracy of the entire system at this time: As shown in, text recognition consists of four parts. Now we can find the system accuracy after optimization for each part. The question is, how can we improve the accuracy of the entire system? We can see from the table that, if we have optimized the text moderation part, the accuracy will be72%Add89%If we optimize the character segmentation, the accuracy is only from89%To90%If character recognition is optimized90%To100%In contr

take an average of this evaluation mode.It is a useful algorithm to use the F-score algorithm to evaluate both precision and recall rates . The PR of the molecule determines that the precision ratio (P) and recall (R) must be large at the same time to ensure that the F score values are larger. If the precision ratio or recall rate is very low, close to 0, the direct result of the PR value is very low, approaching 0, that is, F score is also very low.At this point we compare three algorithms, we

distribution with the mean value of μ 0 and the covariance matrix of Σ, X | y = 1 follows the multivariate Gaussian distribution where the mean value is μ1 and the covariance matrix is Σ (This will be discussed later ). The log function for maximum likelihood estimation is recorded as L (ø, μ 0, μ 1, Σ) = Log 1_mi = 1 p (x (I) | Y (I); μ 0, μ 1, Σ) P (Y (I); ø), our goal is to obtain the parameter ø, μ 0, μ 1, Σ to make L (ø, μ 0, 1, Σ) to obtain the maximum value. The values of the four para

hyper-plane (w,b) and the entire training set is defined as:Similar to the function interval, take the smallest geometric interval in the sample.The maximum interval classifier can be regarded as the predecessor of the support vector machine, and is a learning algorithm, which chooses the specific W and b to maximize the geometrical interval. The maximum classification interval is an optimization problem s

be able to find the global optimal solution.When the training sample is very large, each update parameter needs to traverse all the sample calculation total error, so that the learning speed is too slow; this time the random gradient descent algorithm that calculates the error update parameters of a sample is usually more thanThe batch gradient descent method is faster. (Theoretically, there is no guarantee that the random gradient descent can conver

based on the minimum mean variance. The closer to the predicted point, the heavier the weight, which is to use the points near the check to give higher weights. The most common is the Gaussian nucleus. The weights corresponding to the Gaussian nuclei are as follows:In (Formula 2), the only thing we need to make sure is that it's a user-specified parameter that determines how much weight is given to nearby points.Therefore, as shown in (Equation 3), local weighted linear regression is a non-para

does not introduce a matrix, which is easy to calculate and can be correctly executed if there are few samples. The multi-element model is complex to calculate after the matrix is introduced. to calculate the inverse of the matrix, the model must be executed when the sample value is greater than the feature value. ------------------------------------------Weak split line---------------------------------------------- Although exception detection is mentioned in this article, it is used to in