hi.
I have two data sets, one is normal data and the other is a dataset with abnormal data. The strategy is training an auto-encoder network with normal data and flow data that has abnormal data. This is a basic anomaly detection with auto encoder and there are many examples on the Internet. However, my data has only four variables which means the input layers’ shape is four. Most auto-encoder examples have multiple variables, and the number of nodes decreases as the layer deepens. However, I have never seen an auto-encoder exmple with only four variables. How do I solve the anomalies detection problem with only four variables? I have tried several machine learning algorithms, but they didn’t work well.
Hi @Yong-Hwan_Kim, apart form detecting anomalies auto-encoders are also used for dimensionality reduction tasks (reduce high-dimensional space into a low-dimensional space so that the low-dimensional representation contains some meaningful properties of the original data). If you pass data with low dimensions in the latent space the data will be reduced much and will not get accurate results. Is there any chance to increase the variables? Thank You.