Thanks to visit codestin.com
Credit goes to www.scribd.com

Scribd
Upload
18 views2 pages

Deep Learning Assignment - 1

The document outlines an assignment for a deep learning course, requiring students to create various neural network models for tasks such as house price prediction, image classification, and text sentiment analysis. Students are instructed to utilize different datasets, apply techniques like Stochastic Gradient Descent, and compare model performance based on various hyperparameters and architectures. The assignment emphasizes practical implementation and analysis of model accuracy, regularization techniques, and the impact of different activation functions.

Uploaded by

rsd
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
18 views2 pages

Deep Learning Assignment - 1

The document outlines an assignment for a deep learning course, requiring students to create various neural network models for tasks such as house price prediction, image classification, and text sentiment analysis. Students are instructed to utilize different datasets, apply techniques like Stochastic Gradient Descent, and compare model performance based on various hyperparameters and architectures. The assignment emphasizes practical implementation and analysis of model accuracy, regularization techniques, and the impact of different activation functions.

Uploaded by

rsd
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
You are on page 1/ 2

DEPARTMENT OF ARTIFICIAL INTELLIGENCE AND DATA SCIENCE

ACADEMIC YEAR 2025-2026 ODD


AD3501-DEEPLEARNING
ASSIGNMENT -1

Date of submission :

Assignment Question CO PO

Create a deep neural network program to predict the price of a house CO1 PO1,PO2
based on inputs like the area, number of bedrooms, bathrooms, location ,PO3,PO
score, and house age. Use arrays or matrices to store and process the 5,PO12
input data. Train the model using Stochastic Gradient Descent (SGD) and
test it using a training and validation set. Adjust settings like learning rate
and number of training steps to improve accuracy. Use techniques like
Dropout or L2 regularization to avoid overfitting. Submit both the Python
code and a short report explaining your model, training steps, and how
well your model works.
Use a small neural network on the MNIST dataset. Change CO1 PO1,PO2
hyperparameters like learning rate, batch size, and number of epochs. ,PO3,PO
Record and compare how each change affects the model’s validation 5,PO12
accuracy and loss.
Using the Iris dataset, design a deep feedforward neural network with at CO1 PO1,PO2
least two hidden layers to perform multiclass classification. Train the ,PO3,PO
model and provide the accuracy and confusion matrix. Explain how 5,PO12
softmax is used in the output layer.
Create two deep feedforward models for the same binary classification CO1 PO1,PO2
dataset. ,PO3,PO
5,PO12
 One using ReLU activation
 Another using tanh
Train both and compare their performance. Explain how the choice
of activation function influences the learning process.

Build a CNN model to classify images from a small image dataset (e.g., cats CO2 PO1,PO2
vs. dogs or flowers). Apply data augmentation techniques (like rotation, ,PO3,PO
zoom, flip) and compare accuracy before and after augmentation. 5,PO12
Take a small colored image dataset (e.g., CIFAR-10 or your own collected CO2 PO1,PO2
dataset). Build a CNN model on the original colored images and then train ,PO3,PO
the same model on grayscale versions of the images. Compare the training 5,PO12
accuracy and comment on the impact of color information.
Train a CNN model using a standard dataset (e.g., flowers dataset). Then CO2 PO1,PO2
test the model using 3–5 photos taken by you using your mobile camera. ,PO3,PO
Preprocess the images as needed and analyze how well the model 5,PO12
performs on real-world images
Build a CNN model for image classification using a dataset like Fashion CO2 PO1,PO2
MNIST or CIFAR-10. Train the same model using two different optimizers ,PO3,PO
(e.g., Adam and SGD) and compare their performance based on accuracy 5,PO12
and training time. Explain which optimizer works better and why.
Build a simple RNN model to classify movie reviews from the IMDB dataset CO3 PO1,PO2
,PO3,PO
as positive or negative. Preprocess the text, build the model, and compare 5,PO12
the accuracy before and after applying padding and dropout.
Using a small stock market dataset (e.g., Apple or Google stock prices), CO3 PO1,PO2
build an RNN model to predict the next day’s price. Use features like ,PO3,PO
previous day prices and visualize the predicted vs. actual values on a line 5,PO12
chart.

You might also like