What's new in TPOT
(Tree-based Pipeline Optimization Tool)
Trang Lê
@trang1618
Clean data
Select features
Preprocess features
Construct features
Select classifier
Optimize parameters
Validate model
Raw data
Automate
Typical pipeline
Automate
Open source AutoML tools
Randy Olson
Ryan J. Urbanowicz
Peter C. Andrews Nicole A. Lavender
La Creis Kidd
Jason H. Moore
- DEAP
- Objective:
- maximize pipeline's cross-validation classification/regression performance
- minimize pipeline's complexity
- Pareto front with NSGA-II
TPOT
Weixuan Fu
Entire data set
Entire data set
PCA
Polynomial features
Combine features
Select k best features
Logistic regression
Multiple copies of the data set can enter the pipeline for analysis
Pipeline operators modify the features
Modified data set flows through the pipeline operators
Final classification is performed on the final feature set
an Example "individual"
GP primitive Dataset selector, Feature selector & preprocessor, Supervised classifier/regressor
Individual Sequence of pipeline operators
Population
Generations
Mutation and crossover
(a) insertion mutation
(b) deletion mutation
(c) swap mutation
(d) substitution mutation
(e) crossover
primitives/operators
-
feature selectors
-
feature preprocessors
-
supervised classifiers/regressors
-
feature set selectors (FSS)
parameters
-
generations
-
population_size
-
offspring_size
-
mutation_rate
-
...
-
...
-
template
Template + feature set Selector
Future works
- extend to select 2+ subsets
- re-formulate complexity
- each operator: flexibility (# parameters?), runtime
- number of features used in pipeline
- access over-fitting
- batch sampling, layered TPOT...
- ideas...
- FeatureSetSelector**
Jason Moore
Weixuan Fu