An optimized neural network coding for IoT real time applications to obtain Throughput enhancement, Low latency Implemented using Deep learning & Raspberrypi

first of all, you should make me convenient and comfortable in any clarifications from my side without any hesitations. These are the details of my reference NNC paper : • Tools used Keras and tensor flow of 2.0. It contains three parts encoder, an intermediate node, a decoder; an Autoencoder is used at the encoder, an intermediate node, and a decoder. The monkey image is divided into 4 pieces and sent through a network or ISP. Encoding is done at the intermediate node also. MNIST data set (Handwritten digits from 0 to 9) is used. Encoder decoder works jointly. Finally what happens is low SNR regimes also this system results in better SNR when compared to existing techniques. Requirements: Phase I Project (Total budget:20,000 in Indian Rupees & time line:15 days) • An improvisation of NNC paper (NNC – neural network + network coding) should be implemented in real-time on an IoT device. • What novelty & uniqueness can you propose? • Need to optimize autoencoder by using the application Numerical computing / scientific computing libraries available in Keras and TensorFlow 2.0 in low SNR regimes than that of NNC paper. • *Validate with an existing system • *Visualization using graphs and plots & validate with NNC paper (see fig 4 and fig 5 in NNC paper ). Also, different visualization such as Histogram, Heat map, error Bar, Scatter plot, Need to plot hyperparameter optimization plot, loss vs epoch • *QoS – Signal to Noise Ratio(SNR), Latency, Packet loss, Throughput, CPU processing speed/ computational complexity, Accuracy of the model, Loss function, power consumption, Reliability • *The proposed idea should be demonstrated in real-time for any of IoT applications like Fog Computing, video surveillance, etc., (Tensorflow lite may be convenient for u) • Encoder, Decoder, Intermediate node are the 3 parts of the system. The intermediate node is A NODE IN NETWORK [login to view URL], the end device is IoT. You need to specify which layer in the OSI model we are focusing on with an explanation • *The hardware I have Raspberry pi 3b+. You need to guide me on how to install the necessary library in my raspberry pi to complete the demonstration part of the application from remote access. • The NNC paper they have done only for the MNIST data set. So you have to take the color image, video, Text & sequence data • The NNC paper they took only one type of neural network. So you have to develop 3 different types of Deep learning model like CNN, etc., and Image – CNN Video – Computer vision Text, sequence data – RNN • CPU, GPU, TPU (for ex: Google colab configuration settings) processing time and power consumption comparison should be visualized in a table for ur implementation and validate it with NNC paper. note: *very important • To provide detailed Documentation of the proposed concept in detail with relevant mathematical equations, bcoz linear algebra is the whole theme of my work. • To provide Algorithm of the proposed concept along with necessary flow chart • Should explain the coding thoroughly so that I can understand better & can able to present to my panel. • DL model structure • Encoder working • Intermediate node working • Decoder working Phase II Project (5000 in Indian Rupees timeline: 3 days) Image – CNN Video – Compute vision Text, sequence data - RNN ur Deep learning model should select appropriate Deep Learning model at run time based on the inputs at real-time on IoT device If I am satisfied with ur proposed concept I will hire you.