A mobile robot is equipped with a proximity sensor that has been measured at the factory to offer Gaussian distributed measurements xi, with i being the i-th example index. The measurements help determine via a binary classifier if the robot is close or far from an obstacle. In this use case you can ignore any online setting - assume that the moasurements are stored in a buffer (memory) before attempting any classification. After a year in operation, the sensor is now faulty: intermittently and completely at random and independent of anything else does not report all measurements (missing measurements). A. (10 points) Suggest a method that can impute (provide values) for the missing measurements. B. (10 points) Suggest a probabilistic binary classification model of the posterior y^ that will include a binary random variable ri indicating whether the measurement is observed or not. This means that relative to ( A ) previously the model here knows which measurements are missing. PS: We do not care how to compute such model - all you should care at this point is how to express it. C. (10 points) The vendor has started to receive field reports and telenetry streams of its robots equipped with this sensor and they try to avoid a recall. They need to come up with a new classification model where their Recurrent Neural Network (RNN) classifier can account for the missing measurements. Can you suggest a way that this can be achieved ? PS: (B) and (C) are independent questions - do not answer (B) by considering (C) and vice-vena