1.

Construct a simple neural network performing classification on MNIST Handwritten

Digit dataset

2. Build a CNN and train it with the Labeled Faces in the Wild dataset to determine
how well a CNN can be trained to perform facial recognition.

3. Build a Deep Neural Network for XOR problem using Keras

4. Implement handwritten digit classification using Neural Network

5. Construct a face recognition model using CNN

6. Write a python program to implement sentiment analysis using RNN

7. Write a python program to implement the following task

i.) Load dataset ii.) Perform data preprocessing iii.) build a model iv.) check the
accuracy

8. Perform Language modeling using RNN

9. Implement Machine Translation using Encoder-Decoder model

10. Implement part of speech tagging using sequence to sequence model

11. Write a python program to implement Name Entity Recognition.

12. Write a python program to implement text generation using LSTM

13. Write a python program for generating new set of images from the existing images
using Generative Adversarial Network.

14. Build a simple neural Network to perform image classification

15. Implement image classification using ANN and CNN and compare the accuracy

16. Build a CNN to classify flower images

17. Load cifar10 dataset from tensorflow and perform classification using CNN

18. Implement the following task using python

1. Load IMBD dataset

2. Remove html tags
3. Remove punctuation and numbers
4. Remove single character
5. Filter stopwords
1)MNIST(Modified National Institute of Standards and Technology.) dataset
is a major project made with the help of Neural Network. It basically detects the
scanned images of handwritten digits.

It not only detects scanned images of handwritten digits but also allows writing digits
on the screen with the help of an integrated GUI for recognition.
the primary goal is to train a simple neural network to recognize and classify handwritten digits (0-
9) from the MNIST dataset.

1. Neural Network

 Definition: A computational model inspired by the human brain, consisting of layers

of interconnected nodes (neurons) that process and transform data.
 Uses: Image recognition, speech processing, natural language understanding, etc.
 Types:
o Feedforward Neural Network (FNN)
o Convolutional Neural Network (CNN)
o Recurrent Neural Network (RNN)
o Generative Adversarial Network (GAN)
 Components:
o Input Layer
o Hidden Layers
o Output Layer
o Weights and Biases
o Activation Functions (e.g., ReLU, Sigmoid)

2. MNIST

 Definition: A dataset of 70,000 grayscale images (28x28 pixels) of handwritten digits

(0-9), widely used for training and testing machine learning models.
 Uses: Benchmarking classification algorithms.
 Advantages: Simple, easy-to-use, and well-labeled.
 Disadvantages: Limited complexity (not ideal for advanced tasks).

3. Classification in Deep Learning

 Definition: A supervised learning task where a model assigns input data to one of
predefined classes (e.g., digit classification: 0-9).
 Uses:
o Image classification (e.g., identifying objects).
o Text classification (e.g., spam detection).
o Speech recognition.
 Types:
o Binary Classification (e.g., yes/no).
o Multi-class Classification (e.g., 0-9 digits).
2) 2. Build a CNN and train it with the Labeled Faces in the Wild dataset to determine
how well a CNN can be trained to perform facial recognition.

Goal

The primary goal is to build and train a Convolutional Neural Network (CNN)
using the Labeled Faces in the Wild (LFW) dataset to evaluate how
effectively a CNN can perform facial recognition. This involves classifying
images of faces based on their identities.

Build the CNN Architecture

 Input Layer: Accepts the processed image input (e.g., 64x64x3 for RGB images).
 Convolutional Layers: Extract spatial features such as edges and textures.
 Pooling Layers: Reduce the dimensions and computation (e.g., MaxPooling).
 Dropout Layers: Prevent overfitting by randomly deactivating neurons during
training.
 Flatten Layer: Convert the 2D feature maps into a 1D vector for dense layers.
 Dense (Fully Connected) Layers: Learn complex relationships between features.
 Output Layer: A softmax layer with neurons equal to the number of classes (unique
faces).

Train the CNN

 Define the number of epochs (iterations over the dataset) and batch size.
 Use the training data to optimize the weights of the model.

Evaluate the Model

 Test the trained model on the test set to calculate performance metrics such as:
o Accuracy
o Precision
o Recall
o F1-Score

CNN:

Definition:

A Convolutional Neural Network (CNN) is a type of deep learning model specifically

designed to process structured grid-like data, such as images. It uses convolutional layers to
extract spatial features like edges, textures, and shapes, making it particularly effective for
visual and spatial data analysis.

Uses of CNNs:

Image Processing and Recognition, Natural Language Processing (NLP), Speech

recognition.
Types of CNNs in Deep Learning:

LeNet, alexnet, vggnet, restnet, inception, rcnn(region based cnn)

3) Build a Deep Neural Network for XOR problem using Keras

Goal

The goal of this task is to build and train a Deep Neural Network (DNN) using
Keras to solve the XOR problem,

XOR Truth Table:

Input 1 Input 2 Output

0 0 0

0 1 1

1 0 1

1 1 0

Keras:

Definition:

Keras is an open-source deep learning framework that provides a high-level

interface for building and training deep learning models. It is user-friendly,

Uses of Keras:

1. Model Development:
o Build deep learning models for classification, regression, and
clustering.
2. Rapid Prototyping:
o Allows quick experimentation with neural networks.
3. Transfer Learning:
o Use pre-trained models for applications like image recognition and
NLP.
Types of Models in Keras:

1. Sequential API:
o Simple, linear stack of layers.
2. Functional API:
o For building more complex models with shared layers or multi-
input/output

DNN:

Definition:

A Deep Neural Network (DNN) is a type of artificial neural network with

multiple hidden layers between the input and output layers.

Uses:

1. Image Classification:).
2. Speech Recognition:
3. Natural Language Processing (NLP

Components of DNN:

1. Input Layer:
2. Hidden Layers:
3. Output Layer:
4. Activation Functions

4) Implement handwritten digit classification using Neural Network

Handwritten digit classification is a machine learning task where a model is trained to
recognize and categorize images of handwritten digits (e.g., from 0 to 9) based on patterns
in the data.

Uses:

1. Optical Character Recognition (OCR):

2. Banking Applications:
3. Postal Services:
4. Education:

Same as exp 1 (neural network)

5) 5. Construct a face recognition model using CNN

Same as exp 2

6)write a python program to implement sentiment analysis using RNN

Sentiment analysis:

Definition:

Sentiment analysis is the process of using machine learning or deep

learning techniques to determine the emotional tone (positive,
negative, or neutral) behind a body of text, often used to analyze
opinions or feedback.
Uses:

1. Customer Feedback:
o Analyze reviews on products and services.
2. Social Media Monitoring:
o Gauge public opinion on events or brands.
3. Market Research:
o Understand consumer sentiment for better decision-making.
4. Political Analysis:
o Analyze voter sentiment during elections.

Types:

1. Binary Sentiment Analysis:

o Classifies text as positive or negative.
2. Multi-class Sentiment Analysis:
o Distinguishes multiple sentiment categories (e.g., positive, negative, neutral).
3. Aspect-based Sentiment Analysis:
o Analyzes sentiment for specific aspects (e.g., "camera" in a phone review).

RNN:Recurrent Neural Network (RNN)

Definition:

An RNN is a type of neural network designed to handle sequential data by

maintaining a memory of previous inputs, making it effective for tasks like
time-series analysis and text processing.
Uses:

1. Natural Language Processing (NLP): Language modeling, sentiment analysis, and

machine translation.
2. Speech Processing: Speech-to-text conversion.
3. Time-series Data: Stock price prediction and weather forecasting.
4. Video Analysis: Action recognition in video sequences.

Components:

1. Input Layer: Feeds sequential data into the network.

2. Hidden Layer: Contains recurrent units to capture temporal patterns.
3. Output Layer: Generates predictions (e.g., next word, sentiment).
4. Activation Functions: Typically uses Tanh or Sigmoid for non-linearity.
5. Loss Function: Cross-entropy for classification or MSE for regression tasks

7) i.) Load dataset ii.) Perform data preprocessing iii.) build a model iv.) check the
accuracy

Load Dataset:

 Use libraries like pandas, numpy, or frameworks like TensorFlow/PyTorch to load

the data.

Perform Data Preprocessing:

 Cleaning: Handle missing values, remove duplicates.

 Normalization: Scale data (e.g., Min-Max scaling for numerical features).
 Encoding: Convert categorical data to numerical (e.g., one-hot encoding).
 Splitting: Divide the data into training and testing sets.

Build a Model:

 Define the model architecture (e.g., layers for ANN, CNN, or RNN).
 Compile the model with appropriate loss, optimizer, and metrics.
 Train the model using the prepared dataset.

Check the Accuracy:

 Evaluate the model on the test data.

8. Perform Language modeling using RNN

Language modeling using RNN:

Definition:

Language modeling using RNN involves training a Recurrent Neural Network

(RNN) to predict the next word or sequence of words based on previous words
in a sentence or paragraph.

Goal:

The goal is to develop a model that can predict the probability of the next word
in a sequence, helping in tasks like text generation, machine translation, and
speech recognition.

Components in RNN for Language Modeling:

1. Input Layer: Takes sequences of words or characters as input.

2. RNN Layers: Capture sequential dependencies in the text.
3. Output Layer: Predicts the next word or sequence.
4. Loss Function: Usually, categorical crossentropy for predicting word probabilities.

Rnn(same as exp 6)

9. Implement Machine Translation using Encoder-Decoder model

Machine Translation :

Definition:

Machine translation in deep learning is the process of automatically translating

text from one language to another using neural networks, particularly sequence-
to-sequence models like Encoder-Decoder.

Uses:

1. Language Translation: Convert text from one language to another (e.g., Google
Translate).

Encoder-Decoder:
the encoder, which encodes the input sequence (e.g., a sentence in one language), and the
decoder, which generates the translated output sequence (e.g., the sentence in another
language).

Goal:

The goal is to train a model that can translate text from one language to another
by learning to map an input sequence (source language) to an output sequence
(target language).

Uses:

1. Language Translation: Convert text from one language to another (e.g., English to
French).
2. Text Summarization: Generate concise summaries of longer text.
3. Speech Translation: Convert spoken language in one language to text in another.