Explaining the Challenges of Accountability in Machine Learning Systems Beyond Technical Obstacles

2. Types Of Machine Learning (Ml) Algorithms

ML techniques shown in Figure 1, were broadly defined into various categories like:

• 1.

  Supervised Learning

• 2.

  Unsupervised Learning

• 3.

  Reinforcement Learning

Figure 1.

ML classification

Deep ML models, particularly deep neural networks, help in solving a range of complex applications. The diagnosis of a disease by models using medical images or other medical data is one demonstrative example. At a time, DL mechanisms frequently operate like black-box models, with the operation's specifics frequently remaining largely unknown (Deo, R. C. 2015). In this instance, it’s hard to define how the project for the particular decision succeeded. As a consequence of this, machine learning models have a hard time fitting into a lot of important applications, like medicine, where doctors need to know what a specific diagnosis means to choose the right treatment. Many methods have been developed to define Deep ML algorithms with comprehend the decision-making procedure as shown in Figure 2. These terms are due to the absence of Explaining techniques in various ML models.

Figure 2.

The overall population score for different categories of ML algorithms

The methodologies for elucidating black-box ML techniques were categorized into 2 divisions namely local approaches, which provide explanations based on individual test cases, and overall mechanisms that aim to clarify broader performance in this particular mechanism. Within these explanations, the contribution of each input feature holds paramount significance. It is a fundamental assumption that a comprehensive explanation assigns a numerical value to each feature, denoting its influence overestimation.