代写ENGN4536/6536 Wireless Communications AI Mini Project Report: Deep Learning for PAPR Reduction in

- 首页 >> Algorithm 算法

ENGN4536/6536 Wireless Communications

AI Mini Project Report: Deep Learning for PAPR Reduction in OFDM

Preliminary: Write a brief commentary on the libraries used in this project (5 marks).

1. What is the functionality of numpy library?

2. What is the functionality of keras library? Particularly, what is the functionality of the Adam optimiser?

3. What is the functionality of EarlyStopping?

4. What is the functionality of tensorflow_probability?

5. What is the functionality of matplotlib.pyplot?

Please write a very brief description of the functionality of each library used in this project.

Commentary: (roughly 50-100 words)

Task 1: Define system parameters and generate dataset (5 marks: 2.5 for commentary + 2.5 for source code and output).

Why do we include the following system parameters in the simulation? Below are three exemplary answers.

1. Fading parameter: Fading parameter is used to generate Rayleigh fading realisations.

2. Modulation order: Modulation order is used to identify the number of bits per symbol transmitted in one orthogonal subcarrier in OFDM systems.

3. Learning rate: Learning rate sets the step size of the minimisation problem.

Commentary: (roughly 50-100 words)

1. SNR:

2. Number of subcarriers:

3. Weight parameter:

4. Batchsize:

5. Epoch:

Task 2: Define a constellation mapper model (15 marks: 7.5 for commentary + 7.5 for source code and output).

1. What is the functionality of the Batch Normalization layer?

2. What is the shape of the fifth Dense layer? Why is the ‘tanh’ activation function appropriate for this layer?

3. What is size of the output of the IFFT operation, i.e. how many symbols are generated by the IFFT for being transmitted at one subcarrier?

Commentary: (roughly 100 words)

Task 3: Channel modelling and equalisation (15 marks: 7.5 for commentary + 7.5 for source code).

1. How is the noise generated, e.g. based on the average symbol energy (or average symbol power) and the given SNR value?

2. How is the fading gain generated, e.g. which library is used?

Commentary: (roughly 100 words)

Task 4: Define a constellation demapper model (10 marks: 5 for commentary + 5 for source code and output).

1. What is the shape of the corrupted bit sequence?

2. What is the shape of the final Dense layer?

3. Why is the ‘sigmoid’ activation function appropriate for the final Dense layer?

Commentary: (roughly 100 words)

Task 5: Define and compile the autoencoder model. Test trained autoencoder model’s performance (30 marks: 15 for commentary + 15 for source code and output).

1. Why is the Adam optimiser chosen for the autoencoder?

2. Why is the symbol_papr function appropriate for the constellation mapper?

3. Why is the binary cross entropy function appropriate for the constellation demapper?

4. What is the weight of the loss function for the constellation mapper? Please state the appropriate numerical value if weight is fixed or the parameter if weight is variable.

5. What is the weight of the loss function for the constellation demapper? Please state the appropriate numerical value if weight is fixed or the parameter if weight is variable.

6. What are the two outputs produced by the autoencoder?

7. What is the input to the trained autoencoder?

8. Please write the key steps to compute the average PAPR of OFDM symbols produced by the autoencoder.

Commentary: (roughly 150-200 words)

Task 6: Compute classical OFDM PAPR performance. Plot autoencoder vs classical OFDM PAPR

1. Please write the key steps to compute the average PAPR of OFDM symbols in classical OFDM systems.

2. Please write the key steps to compute the probability values for an OFDM scheme (e.g. the autoencoder-based OFDM scheme and the classical OFDM scheme).

performance (20 marks: 10 for commentary + 10 for source code and output).

Commentary: (roughly 100 words)



站长地图