Please enable JavaScript.
Coggle requires JavaScript to display documents.
Neural Network Learning Rules (Competitive Learning (Steps (1.…
Neural Network
Learning Rules
Supervised
Learning
Regression
Predict new value
based on past / inference
Compute
new values
for
dependent variable
Classification
Divide sample into
classes
Use a trained set of
previously labeled data
Unsupervised
Learning
Clustering
Portion a set of data into
clusters
that
share some
common characteristic
K-Means
Competitive
Learning
Unsupervised Learning
Outputs are in competition
for input patterns
Steps
1. Initialisation
Assign small random value to
weight
[0,1]
Assign a small positive value to
learning rate, α
2. Similarity
Matching
Apply input vector X &
find the
winner-takes-all neuron jx
at iteration p
3. Learning
Update synaptic weight
w(p+1) = W(p) + △W(p)
△W(p) =
- α[x - w(p)] if j ∈ Λj(p)
- 0 if j ∉ Λj(p)
Λj(p) is neighborhood function centered
around the winner-takes-all
neuron jx at iteration p
Reinforcement
Learning
Q-Learning