Uncertainty is the most challenging area in self-driving cars and in solving problems by machine learning algorithms. According to Wikipedia, uncertai

How to Solve Unknown Unknowns with Machine Learning Algorithms

submited by
Style Pass
2021-05-29 03:32:34

Uncertainty is the most challenging area in self-driving cars and in solving problems by machine learning algorithms. According to Wikipedia, uncertainty refers to unknown information and applies to predictions of future events, physical measurements already made, or the unknown. The source of uncertainty can be inherent in the machine learning model or imposed on the system from outside the system, which means that it is from the system environment that shows its effect on the input data to the system.

Self-driving cars must be able to cope with a large amount of uncertainty due to the environment, and due to the limited performance of car sensors, not all of them are recognizable. The only way to find solutions to all possible situations is to learn them step by step like a human. A person who does not know how to drive in snowy weather or on slippery roads is like a self-driving car that has not yet learned this situation.

Environment as a source of uncertainty forces the use of machine learning algorithms to solve such problems. Classical algorithms and software solutions can not cope with unknown issues. A simple example is identifying spam emails. The number of spam detection rules is huge and implementing all the rules in classic system development is not realistic, and the only solution to such problems is machine learning.

Leave a Comment