Fichas
Sin título
Fichas del conjunto (5)
- 1Machine LearningArtificial Neural Networks:
– Neurons: linear threshold units
– Networks:
• Feed forward (Rosenblatt)
• Fully connected graphs (Hebb)
– Learning:
• Supervised (backprop)
• Unsupervised (correlations) - 2ML Internal Aspects • What is changing?– Weights, data structures, executable code
– How are changes made?
• Learning rule:
– Feedback + internal state ==> new state
» Weight adjustments
» Structure modifications
» Code changes - 3Learning Weights:Simplest, most pervasive:
• Artificial neural networks
• Bayesian inference networks
• Parameterized models - 4Learning rule: weight update procedure– Dependent on feedback:
• Indirect: reinforcement
• Direct: error correction
• Embedded domain knowledge: what to change - 5External Aspects• Form of feedback:
– Direct: correct action
– Indirect: reinforcement
• Frequency of feedback
– After each action
– After a sequence of actions
Información
- Descripción
- Intro to AI Fichas sobre Sin título, creado por tifani.obrien el 09/05/2014.
- Creado por
- tifani.obrien
- Visitas
- 7
- Valoración