Analysing Contributions of Components and Factors to Pork Odour Using Structural Learning with Forgetting Method

A novel neural network based approach to analysing contributions of odour components and factors to the perception of pork farm odour is proposed. A multi-component multi-factor odour analysis model is developed and learnt by an algorithm called structural learning with forgetting. Through the learning, unnecessary connections fade away and a skeletal network emerges. By analysing the resulting;skeletal networks significant odour components and factors can, be identified, and thus a more thorough understanding of odour model can be obtained. The proposed approach is tested with a pork farm odour database. The results demonstrate the effectiveness of the proposed approach.