架构

Autoencoder

定义

A neural network trained to compress data into a lower-dimensional latent space and reconstruct it. Autoencoders are used for anomaly detection, dimensionality reduction, and as components in generative models like VAEs.

相关术语

Encoder-Decoder

A two-part neural network architecture where the encoder compresses an input into a latent representation and the decoder generates an output from it. Used in translation, summarisation, and speech recognition systems.

Diffusion Model

A class of generative models that learn to reverse a gradual noise-addition process to produce high-quality images, audio, or video from random noise. Diffusion models power leading image generators such as Stable Diffusion and DALL-E 3.

Unsupervised Learning

A machine learning approach that discovers structure in unlabelled data. Common techniques include clustering (grouping similar items) and dimensionality reduction. It is often used for anomaly detection and exploratory data analysis.

了解术语只是第一步，将其落地应用才是第二步。

预约一次探索通话，探讨这些 AI 概念如何转化到您所在的具体行业与业务挑战中。