Scribd
Upload
14 views7 pages

Deep Learning Simulation Assignment - 1

The document discusses the implementation of neural networks using MATLAB's Neural Network Toolbox for function approximation and predictions. It covers a feedforward neural network for a complex mathematical function, a SISO network for housing price prediction using the Boston Housing dataset, and a MIMO network for predicting environmental parameters using an air quality dataset. Each section includes details on network architecture, training parameters, and conclusions about the networks' performance.

Uploaded by

saaha3130
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
14 views7 pages

Deep Learning Simulation Assignment - 1

The document discusses the implementation of neural networks using MATLAB's Neural Network Toolbox for function approximation and predictions. It covers a feedforward neural network for a complex mathematical function, a SISO network for housing price prediction using the Boston Housing dataset, and a MIMO network for predicting environmental parameters using an air quality dataset. Each section includes details on network architecture, training parameters, and conclusions about the networks' performance.

Uploaded by

saaha3130
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd

FUNCTION APPROXIMATION USING NN TOOLBOX

ABSTRACT:
A feedforward neural network is implemented using MATLAB’s Neural Network Toolbox to
approximate a complex mathematical function:

𝑦 = 𝑠𝑖𝑛(𝑠𝑖𝑛(6𝑥)) + 0.5𝑐𝑜𝑠(𝑡𝑎𝑛(3𝑥)) + 𝑡𝑎𝑛ℎ(0.5𝑥) + 0.2𝑥 2

CODE:
OUTPUT:

CONCLUSION:
The feedforward neural network successfully approximated the nonlinear function.
SISO NEURAL NETWORK SIMULATION

ABSTRACT:
Implementing a Single Input Single Output (SISO) neural network to predict housing prices
using the Boston Housing dataset. The model takes the average number of rooms per dwelling
as input and estimates the corresponding house price. The network is trained with
backpropagation to learn the mapping effectively.
DATASET USED:
The Boston Housing dataset is used and it is acquired from Kaggle. It contains 506 data points.

NETWORK ARCHITECTURE AND TRAINING PARAMETERS:


A 3-layer network is implemented with two hidden layers of 50 and 40 neurons to achieve
optimal output. Training is done with backpropagation for 2000 epochs using a learning rate of
0.09. The training is stopped when the maximum number of epochs is reached.
DATA TRAINING AND TESTING:
A total of 300 data points were selected from the Boston Housing dataset. Out of these, 80%
(240 samples) were used for training the neural network, while the remaining 20% (60 samples)
were used to test the model’s performance.

ERROR:
Error while Testing

Error while using 0.01 learning rate and 4000 epochs

Error while using 0.001 learning rate

CONCLUSION:
The implemented SISO neural network successfully learned the relationship between the
average number of rooms and housing prices in the Boston Housing dataset.
MIMO NEURAL NETWORK SIMULATION

ABSTRACT:
Implementing a Multiple Input Multiple Output (MIMO) neural network to predict
environmental parameters using an air quality dataset. The model takes three input features—
CO concentration, NO₂ concentration, and C₆H₆ concentration—and estimates two output
variables, temperature and relative humidity. The network is trained using backpropagation
with multiple hidden layers to effectively learn the nonlinear relationships between air
pollutants and environmental conditions.
DATASET USED:
The Air Quality dataset is used and it is acquired from Kaggle. It contains 9358 data points.
NETWORK ARCHITECTURE AND TRAINING PARAMETERS:
A 5-layer network is implemented with four hidden layers of 50 ,40, 30, 20 neurons to achieve
optimal output. Training is done with backpropagation for 8000 epochs using a learning rate of
0.0001. The training is stopped when the maximum number of epochs is reached.

DATA TRAINING AND TESTING:


A total of 200 data points were selected from the Air Quality dataset. Out of these, 80% (160
samples) were used for training the neural network, while the remaining 20% (40 samples)
were used to test the model’s performance.
ERROR:
Error while Testing

Error while using 0.0001 learning rate and 4000 epochs

CONCLUSION:
The implemented MIMO neural network successfully learned the relationship between the
inputs and the outputs in the Air Quality dataset.

You might also like