Blandat Flashcards
De olika maskin-inlärningarna:
Supervised learning, unsupervised learning och reinforcement learning
Olika faser
Training phase och Infer Phase
Training phase
Learning from known examples/observations and building ML model
Infer phase
Applying the built model on unseen examples, measuring the performance/ accuracy
Supervised learning
Learn with labelled training set, we need to train the machine.
- Regression
- Classification
(-Artificial Neural Networks)
I vilka problem kan man använda supervised learning?
classification and regression problems
Classification
Decision trees, SVM support vector machine, Others
Artificial Neural Networks
Deep learning, classical neural networks and deep.
Regression
Used for data, linear polynomial
Unsupervised learning
Discovering patterns in unlabelled data, machine is self-sufficient in learning.
-Clustering
Clustering
Put common things, objects in cluster. One is K-means, using distances.
Reinforcement learning
Learning based on feedback or reward, machine has its own rules of hit & trial.
Neural network
Works like a human brain with nodes referred to as neurons. Described with layers. Input layer, hidden layer and output layer. Forward and backward propagation is used to make the network adapt and learn. The output is evaluated, that is called backward propagation. Forward is responsible for making predictions and backward for adjusting the model to reduce the prediction errors. Making the network able to improve it’s performance.
DFS (queue, optimality time space)
LIFO queue
Not optimal, O(b^m) time, O(bm) space
BFS (queue, optimality time space)
FIFO queue
Optimal, O(b^d) time, O(b^d+1)
