1, paste the Python environment and pip list screenshots, look at everyone's readiness. Please will not have the conditions for the development of the reasons and plans.
2, paste the video study notes, requires real, not plagiarism, handwriting can take pictures.
I. Introduction
1, the concept of (04:05): Machine learning is a branch of artificial intelligence. We use a computer to design a system, it can in a certain way to learn the training data provided; the third training session, the system can continue to learn and improve on performance; by learning parameter optimization model can be used predictive output related issues.
2, the connotation and denotation of machine learning (19:00)
(1), machine learning problem can be solved to predict given the data. Such as: data cleaning, feature selection, determining algorithm model, parameter optimization and prediction results.
(2), can not solve the problem of machine learning are: large data storage, parallel computing, and so do a robot.
3, machine learning general process (37:35): collecting data → data → cleaned using the characteristic data modeling → → engineering model.
4, machine learning methods (40:30)
(1), linear regression, rate, Loss (45:40)
(2), EM Code, EM algorithm (46:40)
(3), CMM and the image (49:05)
(4), the convolution of the image (49:35)
(5), to ICA mean separation, the separated signal with noise (50:10)
(6), SVM algorithm (51:37)
(7) Crawler crawling data (52:10)
(8) HMM word (MLE) (52:50)
(9) LDA (53:30)
(10) Other content: maximum entropy, clustering, dimension reduction, SVM, topic model pLSA / LDA, CRFs, variational derivation Variation Inference, deep learning (55:10)
5, mathematics (62:30)
(1) the number of columns is configured (65:40)
(2) Natural parameters (66:30)
(3) the derivative (67:40)
(4) the derivative (70:00) of the common function
(5) Application 1 (71:30)
(6) integration application 2 (77:00)
(7) Taylor official -Maclaurin official (82:20)
(8) Taylor formula is applied (84:50)
(9) direction derivative (97:16)
(10) a gradient (102: 50)
(11) Γ function (111: 10)
(12) convex function (114: 55)
(13) Probability (128: 40)
3. What is machine learning, what classification? With case, write your understanding.
A: Machine learning is a branch of artificial intelligence. We use a computer to design a system, it can in a certain way to learn the training data provided; the third training session, the system can continue to learn and improve on performance; by learning parameter optimization model can be used predictive output related issues.
Machine Learning Category:
(1) supervised learning
Supervised learning is the use of limited training data set has been labeled by some to establish a model of learning strategies / methods to achieve labeling of new data / instances (classification) / mapping. Supervised learning classification and labeling requirements of training samples is known, the more accurate classification and labeling, the more representative sample, the higher the accuracy of the learning model. Supervised learning in natural language processing, information retrieval, text mining, handwriting recognition, spam detection and other fields have been widely applied.
Supervised learning input sample set is labeled classification and labeling, quite literally, given a set of standard answer. Supervised learning from such a given sample classification label of a centralized learning function, when new data arrives, this function can be predicted based on classification and labeling of new data.
In supervised learning, input data is called "training data", each set of training data have a clear identity or results, such as anti-spam system "spam" "Not Spam" classification.
When build predictive models, supervised learning to build a learning process, the predicted results were compared with a "training data" the actual results, continuously adjust predictive models, until the predicted results of the model to achieve a desired accuracy.
The most typical supervised learning algorithms, including regression and classification.
(2) unsupervised learning
Unsupervised learning using limited data hiding unmarked described structure / rule of unlabeled data. Unsupervised learning and training samples do not need to manually labeled data, ease of compressed data storage, reducing the amount of calculation algorithms to enhance the speed, also avoid misclassification problem caused by offset positive and negative samples, mainly for economic forecasting, anomaly detection, data mining , image processing, pattern recognition, such as organizing large-scale computer clusters, social network analysis, market segmentation, astronomical data analysis.
Compared with unsupervised learning supervised learning, no pre-marked sample set a good classification labels, that is not predefined standard answers. It does not tell the computer how to do it, but let their own computer to learn how to classify the data, and then take some form of incentive for those who correctly classified behavior.
In unsupervised learning, the data are not specifically identified, learning model is to infer some of the internal structure of the data. Common scenarios include association rule learning, and clustering. Common algorithms include Apriori algorithm, KMeans algorithm, a random forest (random forest), principal component analysis (principal component analysis) and so on.
(3) semi-supervised learning
Semi-supervised learning between supervised learning and unsupervised learning, the main problem is to use a small amount of labeled samples and a large number of unlabeled samples for training and classification, so as to reduce the cost of labeling, the purpose of improving learning ability.
In this learning mode, the input data is identified section, part is not identified, this learning model can be used to predict, but this model, to learn the internal data structures to organize data in order to reasonably predict.
Application scenarios include classification and regression algorithms including some commonly used extension of supervised learning algorithms, these algorithms are not identified on the first attempt to model data, and then identify the data to predict on this basis. As shown on the reasoning (graph inference) algorithm Laplace or SVM (Laplacian SVM) and so on.
(4) reinforcement learning
Reinforcement learning is an intelligent system mapping behavior from the environment to learn in order to strengthen the function of the signal value of the maximum. As the external environment provided little information, learning reinforcement learning system must rely on their own experience.
Strengthen the goal of learning is to learn from the environment mapping state to act, so that the behavior of the agent selected for its ability to get the most reward environment, so that the external environment was evaluated as the best in some sense the learning system. In its robot control, unmanned, chess, industrial control and other fields successfully applied.
In this learning mode, the input data as feedback to the model, unlike the supervision model as input data only as a model to check wrong way. In reinforcement learning, input data directly back to the model, the model must make adjustments immediately. Common scenarios include dynamic systems and robot control.
Common algorithms include Q-Learning and learning time difference (temporal difference learning).
Reference website: https://blog.csdn.net/zw0Pi8G5C1x/article/details/89173508