Model Tuning
If there is a need to create a more accurate model particularly while up against benchmark problems there are techniques such as retraining an already trained model, searching for a suitable learning rate using the build-in learning rate finder which is novel to fast.ai. The technique to further tune the model can be found in the below code blocks:
Refer to #CODE BLOCKS 11,12,13 # on Jupyter Notebook
Then using that learning rate you can fit other cycles until the model gets to a suitable accuracy - in our case we are able to get to an accuracy of 89.47%, or an error rate of 0.1052 even without resulting to the next tuning technique by data exploration.
Model Tuning by Data Exploration
Now that we have a model it would be good to look at the images that might be causing the model to be less accurate. Maybe in the process of inputting your data, there might have been some mixed up labels or misclassified images. The first step is to draw a confusion matrix of the species to see the map of the true positives and true negatives and understand which species are making your model less accurate \cite{Ting2017}. In our case, and according to figure \ref{194107}, we have correctly classified 13 of the Pisaster ochraceus and misclassified one as Solaster dawsonii, correctly classified 24 Pycnopodia helianthiodes and misclassified three as two Pisaster ochraceus and one Solaster dawsonii, finally our model has correctly classified 14 Solaster dawsonii and misclassified two as Pycnopodia helianthiodes.
Also, the deeper blue colors can help you identify class imbalance issues, where one species had more data compared to another, giving you insights on whether your model might be struggling with skewed imbalanced data among the classes.