Skip to content

Latest commit

 

History

History
29 lines (20 loc) · 2.7 KB

Deep Learning.md

File metadata and controls

29 lines (20 loc) · 2.7 KB

#dl Deep Learning is a subfield of machine learning concerned with algorithms inspired by the structure and function of the brain called artificial neural networks. It is characterized by its capacity to learn from a large amount of data and capture complex patterns. Here are some key concepts that define how deep learning is different:

Hierarchical Compositionality

  • Cascade of Non-linear Transformations: Deep learning models consist of multiple layers that apply non-linear transformations to the data. These transformations allow the model to learn complex functions mapping inputs to outputs.

  • Multiple Layers of Representations: In deep learning architectures, each layer captures a level of representation of the data. The first layers may capture basic features like edges and curves in images, while deeper layers can capture high-level concepts like objects or even sentiments in text.

End-to-End Learning

  • Learning Goal-driven Representations: Unlike traditional machine learning models where features must be hand-engineered, deep learning models are designed to learn the most predictive features directly from the data, guided by the learning task.

  • Learning Feature Extraction: Deep learning automates the feature extraction process, allowing the model to adapt its features specifically for the task at hand, which often results in superior performance compared to models with hand-crafted features.

Distributed Representations

  • No Single Neuron "Encodes" Everything: In deep learning, information is represented by the activation of a population of neurons rather than a single neuron. This distributed nature of information representation leads to more robust learning.

  • Groups of Neurons Work Together: Deep learning relies on the collective behavior of groups of neurons to represent complex data. This collaborative aspect of neural activity allows deep learning models to be more expressive and to capture intricate patterns in the data.

Why is Deep Learning Powerful?

Deep learning is powerful due to its ability to:

  • Automatically and Hierarchically Learn Features: From simple to complex, without human intervention in the feature design.
  • Model High-level Abstractions: By using multiple processing layers with complex structures or otherwise composed of multiple non-linear transformations.
  • Leverage Large Amounts of Data: Deep learning algorithms are scalable and improve as the size of your data increases.
  • Achieve State-of-the-Art Results: In many domains such as vision, language, and speech, deep learning has set new records for accuracy.

References

#TODO