Original article was published by Seyma Tas on Artificial Intelligence on Medium
Human Touch In Artificial Intelligence
While I was making(creating) machine learning models, I always wondered how effective data cleaning, preprocessing, exploratory data analysis, and feature engineering can transform the simplest models into powerful predictors. Choosing the proper model is crucial but providing the right data to the model is essential. Therefore, the principle of “garbage in, garbage out” is still valid even if the best convolutional neural network is established.
If human touch can make a quite simple model highly accurate, why don’t humans contribute to the training process? Why doesn’t human intelligence boost artificial intelligence in the modeling? Can people help computers before and during training? What if we can draw features, classifiers, outliers while modeling?
Rule-based Systems vs. Algorithm-based Systems
Human learn library
On October 8, 2020, Vincent D. Warmerdam wrote and introduced Human-learn library which makes it easier to create rule-based systems. The main idea is to be more familiar with the data(extended EDA) and to contribute to the model by drawing manually.