Convolutional Deep Learning Neural Network - Consulting

The price offered is for a consulting task. If succesfull it is expected to lead to further work installing and configuring suitable software, with the possibility of an on-going relationship with our company aimed at integrating the use of neural network technologies with our existing systems. We believe that this part (the consulting) will take between 1 to 4 days to complete. However we are ready to discuss this with qualified candidates. The consulting task is to understand our short-term aims, and define how to choose, install and configure a Convolutional (and a non-convo) Deep Learning Network package (such as that offered by Mathworks Matlab or Peltarion), such that it is ready to be trained on our time series data. This is because there are 2 principle ways to use our data; 1) use it in the traditional way, by performing a range of statistical analyses and regressions on it in order to classify the target structures, or 2) treat it as visual image data, with the same goal. In short, our aim is to have specific structures that we know exist in our data classified by the CNN. Once trained, we want it to correctly identify >90% of the instances of these structures in our out of sample data. Meeting this aim will be seen as a very successful outcome. Each time series data set stretches over up to 20 years, and has between 1 and 4 synchronised but different data types. We do not know if inclusion of all the data types will create a more accurate classification of the structures or not; it could be that the accuracy of the classification is best when trained on a single data-type. In any event, to human eyes the different data types are part of a single interrelated dataset, which should be seen as different aspects of a single image. For commercial reasons I cannot post the data here, but I am able to discuss it in detail with individual candidates. The end result of this project should be a working document that specifies a particular Covolutional Deep Learning Network software, along with a quotation to install and configure that software ready to use for our purposes.