Ada.py

Upload of the Statistical Methods for Machine Learning exam project on Adaboost development from scratch done in 2019.

Requirements

This software requires Python3, available here.

Type in bash the following code:

python .\Adaboost.py T

*It can be python3 based on the environment variable.

The parameter T indicates the number of decision stumps to be used. These are generated automatically in the code from the dataset.

One possible execution is as follows:

python .\Adaboost.py 200

Results:

Class Errors \ T	200	400	600	800
Class 1 Train Error	0.09947089947089947	0.09753086419753086	0.09753086419753086	0.09611992945326278
Class 1 Test Error	0.09947089947089947	0.09841269841269841	0.0992063492063492	0.09841269841269841
Class 2 Train Error	0.11190476190476191	0.11031746031746031	0.10864197530864197	0.10767195767195767
Class 2 Test Error	0.11296296296296296	0.11481481481481481	0.11481481481481481	0.1164021164021164
Class 3 Train Error	0.10564373897707231	0.10220458553791888	0.10291005291005291	0.1019400352733686
Class 3 Test Error	0.11587301587301588	0.11507936507936507	0.11587301587301588	0.1164021164021164
Class 4 Train Error	0.026102292768959437	0.024074074074074074	0.023192239858906526	0.022663139329805997
Class 4 Test Error	0.02857142857142857	0.0291005291005291	0.028835978835978836	0.029365079365079365
Class 5 Train Error	0.06499118165784833	0.06031746031746032	0.059171075837742504	0.05793650793650794
Class 5 Test Error	0.07116402116402117	0.06428571428571428	0.06375661375661376	0.06296296296296296
Class 6 Train Error	0.09876543209876543	0.09559082892416226	0.09329805996472663	0.09074074074074075
Class 6 Test Error	0.11084656084656085	0.10185185185185185	0.10264550264550265	0.10238095238095238
Class 7 Train Error	0.02962962962962963	0.029012345679012345	0.02857142857142857	0.027777777777777776
Class 7 Test Error	0.03333333333333333	0.03306878306878307	0.0328042328042328	0.0328042328042328