Shulou(Shulou.com)06/02 Report--

Algorithm is one of the core fields of artificial intelligence (AI).

This paper collates 39 terms commonly used in the field of algorithms, hoping to help you understand this subject better.

1. Attention mechanism

The essence of Attention is to focus from the whole to the focus. Focus limited attention on key information, so as to save resources and get the most effective information quickly.

2. Encoder-Decoder and Seq2Seq

Encoder-Decoder model is mainly a concept in the field of NLP. It does not specifically refer to a specific algorithm, but a general term for a class of algorithms. Encoder-Decoder is a general framework under which different algorithms can be used to solve different tasks.

The framework of Encoder-Decoder well illustrates the core idea of machine learning:

The practical problems are transformed into mathematical problems, and the practical problems are solved by solving mathematical problems.

Seq2Seq (short for Sequence-to-Sequence), literally, enters one sequence and outputs another. The most important thing about this structure is that the length of the input sequence and the output sequence is variable.

3. Q-Learning

Q learning is a value-based learning algorithm in reinforcement learning.

4. AdaBoost

AdaBoost is the abbreviation of Adaptive Boosting.

AdaBoost is an iterative algorithm, its core idea is to train different classifiers (weak classifiers) for the same training set, and then combine these weak classifiers to form a stronger final classifier (strong classifier).

5. Random forest (Random Forest)

Random forest belongs to the method of Bagging (Bootstrap Aggregation for short) in ensemble learning.

In machine learning, a random forest is a classifier that contains multiple decision trees, and its output class is determined by the number of categories output by individual trees.

Four steps of constructing Random Forest

☆ random forests can be used in many places:

Classification of discrete values

Regression of continuous values

Unsupervised learning clustering

Anomaly detection.

6. Learning vector quantization (Learning Vector Quantization)

In computer science, learning vector quantization (LVQ) is a prototype-based supervised learning statistics classification algorithm. LVQ is the supervised version of vector quantization.

Learning vector quantization (Learning Vector Quantization, referred to as LVQ) belongs to prototype clustering, that is, trying to find a group of prototype vectors to cluster, each prototype vector represents a cluster, and the space is divided into several clusters, so for any sample, it can be classified into its nearest cluster. The difference is that LVQ assumes that data samples have category markers, so these category markers can be used to assist clustering.

7. K neighbor (K-Nearest Neighbors)

K-nearest neighbor (K-Nearest Neighbor,KNN) classification algorithm is not only a mature method in theory, but also one of the simplest machine learning algorithms.

The idea of this method is: in the feature space, if most of the K nearest samples (that is, the nearest samples in the feature space) near a sample belong to a certain category, then the sample also belongs to this category.

Officially, the so-called K-nearest neighbor algorithm means that given a training data set, for a new input instance, find the K instances closest to the instance (that is, the K neighbors mentioned above) in the training data set. Most of these K instances belong to a certain class, and the input instance is classified into this class.

Advantages of ☆:

The theory is mature and the thought is simple. It can be used for both classification and regression.

It can be used for nonlinear classification.

The training time complexity is O (n).

No assumptions about data, high accuracy, insensitive to outlier

KNN is an online technology where new data can be added directly to the dataset without having to be retrained.

KNN is simple in theory and easy to implement.

Disadvantages of ☆:

The problem of sample imbalance (that is, the number of samples in some categories is large, while the number of others is very small) is poor.

Need a lot of memory

For data sets with large sample size, the amount of calculation is relatively large (reflected in distance calculation).

When the sample is unbalanced, the prediction deviation is relatively large. For example, there are fewer samples in one category and more samples in other categories.

KNN will perform a new global operation for each classification.

The selection of K value is not the best in theory, and it is often combined with K-fold cross-validation to get the optimal K value.

8. Linear discriminant analysis (Linear Discriminant Analysis)

Linear Discriminant Analysis (LDA) is an induction of Fisher's linear discriminant method, which uses statistics, pattern recognition and machine learning methods to find a linear combination of the characteristics of two kinds of objects or events in order to characterize or distinguish them. The resulting combination can be used as a linear classifier or, more commonly, for dimensionality reduction for subsequent classifications.

Used in the field of facial recognition (in computerized face recognition, each face is expressed by a large number of pixel values. The main role of LDA here is to reduce the number of features to a manageable number before classifying them. Each new dimension is a linear combination of pixel values in the template. The linear combination obtained by Fischer linear discrimination is called Fisher face, and the principal component analysis is called eigenface.)

9. TF-IDF

To put it simply, the vector space model hopes to form the query keywords and documents into vectors, and then use the operations between vectors to further express the relationship between vectors. For example, a more common operation is to calculate the "correlation" between the vector corresponding to the query keyword and the vector corresponding to the document.

10. Meta-learning (Meta Learning)

The idea of meta-learning is the process of learning "learning (training)".

Meta-learning is a relatively new direction in the field of artificial intelligence and is considered to be the key to the realization of general artificial intelligence.

The core of meta-learning is the ability of self-study.

Meta-learning is usually used in optimizing hyperparameters and neural networks, exploring good network structure, small sample image recognition and fast reinforcement learning.

11. Genetic algorithm (Genetic Algorithm | GA)

Genetic algorithm (Genetic Algorithm) is a computational model that simulates the natural selection and genetic mechanism of Darwinian biological evolution. It is a method to search the optimal solution by simulating the natural evolution process.

Genetic algorithms start with a population (population) that represents the potential solution set of the problem, while a population is composed of a certain number of individuals (individual) encoded by genes (gene).

Genetic algorithm draws lessons from the genetic principle in biology. It is a computational model that simulates the biological evolution process of natural selection and genetic mechanism of Darwinian biological evolution theory. It is a method to search the optimal solution by simulating the natural evolution process.

In essence, it is an efficient, parallel and global search method, which can automatically obtain and accumulate knowledge about the search space in the search process, and adaptively control the search process to obtain the best solution.

twelve。 Discriminant model (Discriminative Model)

Discriminant model is a method to model the relationship between unknown data y and known data x in the field of machine learning. Discriminant model is a method based on probability theory. Given the input variable x, the discriminant model predicts y by constructing the conditional probability distribution P (y | x).

13. Production model (Generative Model)

In machine learning, the generation model can be used to model data directly (for example, to sample data according to the probability density function of a variable), or to establish the conditional probability distribution between variables. The conditional probability distribution can be formed by the generation model according to Bayesian theorem.

14. Latent Dirichlet Allocation | LDA

LDA (Latent Dirichlet Allocation) is a document topic generation model, also known as a three-layer Bayesian probability model, which includes three layers of word, topic and document structure.

LDA is an unsupervised machine learning technology that can be used to identify topic information hidden in large document sets (document collection) or corpus (corpus).

15. Heuristic algorithm (Heuristic)

An algorithm based on intuitive or empirical construction gives a feasible solution of each instance of the combinatorial optimization problem at an acceptable cost (computing time and space). The degree of deviation between the feasible solution and the optimal solution can not be predicted.

At this stage, the heuristic algorithm is mainly based on the natural body algorithm, including ant colony algorithm, simulated annealing, neural network and so on.

16. Particle Swarm Optimization (Particle Swarm Optimization | PSO)

PSO algorithm is a kind of evolutionary algorithm, which is similar to simulated annealing algorithm. It also starts from the random solution and iterates to find the optimal solution. It also evaluates the quality of the solution by fitness, but it is simpler than the rules of genetic algorithm. It does not have the "Crossover" and "Mutation" operation of genetic algorithm. It finds the global optimization by following the optimal value currently searched.

17. Deep learning (Deep Learning)

Deep learning is a method based on representation of data in machine learning.

It belongs to the category of machine learning, which can be said to be an upgrade based on the traditional neural network, which is about equal to the neural network. Its advantage is to use unsupervised or semi-supervised feature learning and hierarchical feature extraction efficient algorithm to replace manual feature acquisition.

Deep learning is a new field of machine learning, its motivation is to establish and simulate the human brain for analytical learning neural network, which imitates the mechanism of the human brain to interpret data, such as images, sounds and texts.

18. Artificial neural network (Artificial Neural Network | ANN)

ANN abstracts the human brain neural network from the point of view of information processing, establishes a simple model, and forms different networks according to different connection modes.

To be continued.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.