iteration 0 loss 1.0987398093279945
iteration 100 loss 0.7129422012608996
iteration 200 loss 0.6887726930266873
iteration 300 loss 0.5436778425377103
iteration 400 loss 0.3427991213694607
iteration 500 loss 0.2921075022407062
iteration 600 loss 0.2590305040238777
iteration 700 loss 0.18236065020429468
iteration 800 loss 0.20049390033740888
iteration 900 loss 0.1258016234147451
iteration 1000 loss 0.1077102236889968
iteration 1100 loss 0.09524374700271392
iteration 1200 loss 0.08570292680104846
iteration 1300 loss 0.07833028635138949
iteration 1400 loss 0.07244152243591633
iteration 1500 loss 0.06753888969454228
iteration 1600 loss 0.06342742576765745
iteration 1700 loss 0.05992692038834323
iteration 1800 loss 0.05688405519515976
iteration 1900 loss 0.05419281310098801
iteration 2000 loss 0.05183075921618167
iteration 2100 loss 0.04966551995852915
iteration 2200 loss 0.04772533072984744
iteration 2300 loss 0.045992529755421833
iteration 2400 loss 0.04443767350873418
iteration 2500 loss 0.04300507373620072
iteration 2600 loss 0.04169778815729772
iteration 2700 loss 0.040501514675789135
iteration 2800 loss 0.03940982830968154
iteration 2900 loss 0.03840483967389475
training accuracy: 0.9933333333333333