Final Review
AU STAT627
Emil Hvitfeldt
2021-06-21
1 / 13
`\require{color}\definecolor{orange}{rgb}{1, 0.603921568627451, 0.301960784313725}
\require{color}\definecolor{blue}{rgb}{0.301960784313725, 0.580392156862745, 1}
\require{color}\definecolor{pink}{rgb}{0.976470588235294, 0.301960784313725, 1}`
By @ChelseaParlett
2 / 13
Unsupervised Learning
dimensionality reduction
Clustering
3 / 13
Dimensionality Reduction
We looked at
- Principle Component Analysis
Other methods to consider
- t-distributed Stochastic Neighbor Embedding (t-SNE)
- Autoencoder
- UMAP
- NNMF
4 / 13
Dimensionality Reduction
We looked at
- Principle Component Analysis
- Lasso
Not just useful in linear models
5 / 13
Dimensionality Reduction
We looked at
- Principle Component Analysis
- Lasso
- LDA
6 / 13
Clustering
We looked at
- K-means
- Hierarchical Clustering
Other models
- DBSCAN
- Gaussian Mixtures
7 / 13
Regularisation Methods
- Ridge
- Lasso
Serves different tasks. Can be combined in some cases
8 / 13
Supervised Learning
Regression
Classification
9 / 13
Supervised Learning
Many of the methods we looked at in this class can be used for both regression and classification
We mainly work with 2 types of trade-offs
- Flexibility / Interpretability
- Bias / Variance
10 / 13
Supervised Learning
The models we saw in this class lays the foundation for most models which doesn't go under the neural network umbrella
- xgboost
- lightgbm
- catboost
- stacking
11 / 13
Other considerations
- Implementation and run time
- What metrics are important
- Your problem statement
12 / 13
Thank you!
13 / 13