Deep learning task

To generate synthetic sensor dataset using a modified GAN, cluster the data using ensemble of three clustering algorithms and train models using CNN, CNN-LSTM or Bi-LSTM



My project description: I want to generate synthetic data for human activity recognition with GANs, but the Generator will have signature of path and activity labels (open source python signatory, which supports PyTorch can be used as signature of path). I will attach some papers and architectures so you can understand better. After this, the generated data and the real data's label will be removed, features will be extracted using CNN, then the extracted features will be clustered using an ensemble of k-means, SOM and GMM ( Depends on your choice of clustering algorithms). Then the clusters will be used to train deep learning models. The dataset proposed is available online at https://archive.ics.uci.edu/ml/datasets/PAMAP2+Physical+Activity+Monitoring. We might not consider all the activities. Relevant papers and the overall framework is attached. If its something you are willing to work on, let me know how much you are willing to charge. Remain blessed brother