Tensor data structure¶
Tensor (張量) is the fundamental data structure used to represent information in this context, conceptualized as a multi-dimensional array (多維數組)^[600-developer__big-data__ai__tensorflow-01.md].
Characteristics¶
In the framework of deep learning, the Tensor functions as the primary vessel for data handling within computational models.^[600-developer__big-data__ai__tensorflow-01.md] It serves as the core representation for the complex datasets used in [[neural-networks|Neural Networks]].
The structure is defined by two main components: * Tensor: Represents the data structure itself, specifically a multi-dimensional array^[600-developer__big-data__ai__tensorflow-01.md]. * Flow: Represents the computational model or the flow of data through the system^[600-developer__big-data__ai__tensorflow-01.md].
Application in Neural Networks¶
Tensors are utilized throughout the lifecycle of a [[neural-networks|Neural Network]]^[600-developer__big-data__ai__tensorflow-01.md]: 1. Feature Extraction: Acting as the container for extracted entity features^[600-developer__big-data__ai__tensorflow-01.md]. 2. Structure Definition: Used to define the parameters and architecture of the network^[600-developer__big-data__ai__tensorflow-01.md]. 3. Training: Manipulated during the model training phase to adjust parameters^[600-developer__big-data__ai__tensorflow-01.md]. 4. Prediction: Used to store and process input data for generating predictions^[600-developer__big-data__ai__tensorflow-01.md].
Related Concepts¶
- [[Deep Learning]]
- [[Neural Networks]]