Please enable JavaScript.
Coggle requires JavaScript to display documents.
Machine Learning, Unsupervised Program, Supervised Program, Re-enforcement…
Machine Learning
Supervised
Training Set
Inputs + Outputs
labelled items
Test Set
inputs
non-labelled items
Regression
numerical output
confidence level
continous variable
training set
test set
Classification
category
validation set
class
process
identify variables
identify features
perform classification
Algorithm map
Output Variable
Input Variable
structured data
facial recognition (via deep learning)
Re-enforcement
No Training Data
Trial end error
learning by doing
explore
knowledge of previous outcomes
exploit
3 optimal conditions :check:
complex environment
no training data
continuous learning is possible (and safe)
gaming or drug discovery
uncertain chaotic environments
bad conditions :red_cross:
unable to fail - e.g. autonomous vehicles
Protein Folding :check:
Unsupervised
training data
input variables only
not labelled
unstructured data
Unsupervised Program
learns patterns
measures distance
k-means
Bayesian classifiers
Clustering
groups
Category
Continous variable
accuracy
Supervised Program
answers
new rules
accuracy
Re-enforcement Program
Learns
Training Data is outputted
Innovative Solutions :star:
clear outcome needed :warning:
feedback
Unstructured Data
Clustering :recycle:
Structured Data