Wiki

Receiver Operating Characteristic (ROC) is a test indicator of a system matching algorithm. It is a relationship between the matching score threshold, false positive rate, and rejection rate. It reflects the balance between the rejection rate and false positive rate of the recognition algorithm at different thresholds.

Restricted Boltzmann machine is a kind of random neural network model with two-layer structure, symmetrical connection and no self-feedback.

Simultaneous Localization and Mapping (SLAM) is a technique used in robotics.

Statistical learning is a discipline that builds probabilistic statistical models based on data to predict and analyze data, also known as statistical machine learning.

The alternative loss function is a function used when the original loss function is inconvenient to calculate.

Upsampling, or image interpolation, is mainly used to enlarge the original image so that it can be displayed on a higher resolution display device.

The vanishing gradient problem is a problem encountered when training artificial neural networks using gradient descent and backpropagation.

T-Distributed Stochastic Neighbor Embedding (t-SNE) is a machine learning method for dimensionality reduction.

Treebank is a deep-processed corpus that performs word segmentation, part-of-speech tagging, and syntactic structure relationship tagging on sentences.

Turing machine, also known as deterministic Turing machine, is an abstract computing model proposed by Alan Turing in 1936. Its more abstract meaning is a mathematical logic machine, which can be regarded as the ultimate powerful logic machine equivalent to any finite logical mathematical process.

Specialization is a process from general to specific

A synonym set is a collection of words with the same meaning.

The time step defines how small the time intervals between physics simulations are. In a game engine, this reflects how often a function needs to run.

Parameter adjustment refers to the act of adjusting parameters in order to obtain better results.

Numerical attributes are a type of attribute that quantitatively describes data, meaning that the data is a measurable quantity.

General artificial intelligence refers to an intelligent entity with the same or superior capabilities as humans. It is also called strong artificial intelligence, which can display all intelligent behaviors of normal humans. In order to distinguish it from the AI term of traditional artificial intelligence or mainstream artificial intelligence, a general prefix is added.

Attribute space: The space formed by attributes is also called "sample space" or "input space". Feature space: The attribute space formed by the selected attributes after excluding linear correlation and attributes that are not beneficial to model construction is called feature space. Related concepts Data set […]

The naive Bayes classifier uses the "attribute conditional independence assumption": for known categories, it is assumed that all attributes are independent of each other. Improved naive Bayes: In order to prevent the information carried by other attributes from being "erased" by attribute values that have never appeared in the training set, "smoothing" is usually performed when estimating probability values, and the "Laplace correction" is often used; for […]

Generative adversarial networks are an unsupervised learning method that is implemented by letting two neural networks compete with each other. This method was proposed by Ian Goodfellow in 2014. Generative adversarial networks include a generative network and a discriminative network. The generative network takes random samples in the latent space as input, and the output needs to imitate the training […]

In machine learning, generative models can be used to directly model data or to establish conditional probability distributions between variables. Conditional probability distributions can be based on generative models of Bayes’ theorem. Generative models are suitable for unsupervised tasks such as classification and clustering. Typical generative models include the following: Gaussian mixture models and other mixed […]

Graph theory is a branch of combinatorial mathematics. Its main research object is graph. Graph here refers to a figure composed of several given vertices and edges connecting two vertices. It is often used to describe special relationships between things, where vertices represent things and edges represent the connection between them. Graph theory originated from the Königsberg Seven Bridges Problem, which was solved by Euler in 1736. […]

The gradient explosion problem usually occurs in deep networks and when the weight initialization value is too large. It usually becomes more obvious as the number of network layers increases. By taking the derivative of the activation function, if the result is greater than 1, then as the number of layers increases, the final gradient update will increase exponentially, i.e., a gradient explosion occurs; if the result is less than 1, then the number of layers […]

Eigendecomposition is a method of representing the product of matrices by decomposing them into eigenvalues and eigenvectors. However, only diagonal matrices can be eigendecomposed. Matrix multiplication corresponds to a transformation, which is to transform any vector into a new vector of another direction and length. In this process, the original vector will be rotated and stretched.

An algorithm is an efficient way of representing a finite-length list. In mathematics and computer science, an algorithm can be considered as any well-defined sequence of specific computational steps.