Machine Learning (Machine Learning)
Definition:
Is the research question the model assumptions, using computer learning from training data obtained in the model parameters. And finally the data subject a forecast and analysis.
The basic binary classification problem solving ideas:
(1) establish a hypothetical model (understood as a function). I.e., to establish a straight line, the expression of y = f (x) = W1x + W0.
By determining the value of the model parameter W1 and W0, you can determine the specific function of position.
(2) training model. Use has made it clear classification results of sample data sets (called the training set) for training, namely solving process function parameters. Parameter W0 W1 and get the final value, the resulting final training model.
(3) classified and identified. Using the trained model data to be grouped with discrimination, which is derived which category.
Multi-class classification problem can be divided into two types of classification problem solving. In addition to the linear curve may also be classified, the better. In high dimensional space can be used to classify the surface, but a similar process.
Deep learning (Deep Learning)
Definition:
a technology of machine learning implementation, it is an important branch of machine learning.
Study, depth of learning from model structure artificial neural network is a neural network with multi-hidden layer.
More abstract level features formed by combining low-level features.
Learn a machine learning by example, a simple linear classifier learning depth differences:
Simple classifier is weighted by means of weights w1, w2, w3, w4 x weight of each component.
By linear model y = f (w1x1 + w2x2 + w3x3 + w4x4), if the result is greater than the brackets equal to the threshold, the output 1, output 0 otherwise.
If the depth of the same problem by learning, an intermediate linear model is usually added to the intermediate layer. Wherein U represents a parameter from the input layer to the junction between the intermediate layer wiring right node weight.
Classifier similar linear equation, but this feature is through the intermediate layer, wherein the bottom layer can form an effective combination level features, reducing the number of eigenvalues.
Deep learning often comprises a plurality of intermediate layers (hidden layer), and the network structure is more complicated.
VS depth learning machine learning
machine learning need to manually select the feature
depth study will learn a useful feature
----------------
Disclaimer: This article is CSDN blogger "Tong Lu 0" original articles follow CC 4.0 BY-SA copyright agreement, reproduced, please attach the original source link and this statement.
Original link: https: //blog.csdn.net/qq_44704609/article/details/104374600
