Download as pdf or txt
Download as pdf or txt
You are on page 1of 4

AI LAB REPORT

MNIST Digit Recognition Dataset:


Description:
The MNIST dataset is a classic benchmark dataset in machine learning and computer
vision. It consists of a collection of 28x28 pixel grayscale images of handwritten digits (0
through 9). Each image is labeled with the corresponding digit it represents.

Size:
The dataset contains 60,000 training images and 10,000 test images.

Usage:
MNIST is commonly used as a benchmark dataset for developing and testing algorithms
for image classification tasks, particularly for handwritten digit recognition.

Applications:
It has been widely used for training and evaluating various machine learning models,
including traditional classifiers, neural networks, and deep learning models.

Comparison of Single layer network, Multi-Layer Perceptron, Fully


Connected DNN:
Single Layer Network:
Accuracy: 97.5%

Precision: 98%

Recall: 97%

Comments: The Single Layer Network achieved a high level of accuracy, precision, and
recall on the MNIST digit recognition dataset. However, its performance is slightly lower
compared to the more complex models.

Multi-Layer Perceptron (MLP):


Accuracy: 97.6%

Precision: 98%
Recall: 98%

Comments: The MLP outperformed the Single Layer Network, achieving higher accuracy,
precision, and recall. Its deeper architecture allows it to capture more complex patterns in
the data, resulting in improved performance.

Fully Connected DNN:


Accuracy: 98.1%

Precision: 98%

Recall: 98%

Comments: Similar to the MLP, the Fully Connected DNN achieved superior performance
compared to the Single Layer Network. Its deeper architecture enables it to learn intricate
features from the data, leading to higher accuracy, precision, and recall.

MNIST Fashion Dataset


Description:
The Fashion-MNIST dataset is a variation of the original MNIST dataset, created to serve as
a more challenging alternative. Instead of handwritten digits, it consists of a collection of
28x28 pixel grayscale images of various fashion items such as clothing, footwear, and
accessories. Each image is labeled with the category of the fashion item it represents.

Size:
Similar to MNIST, the Fashion-MNIST dataset contains 60,000 training images and 10,000
test images.

Usage:
Fashion-MNIST is used for the same purposes as MNIST, but it provides a more diverse and
challenging set of images for testing algorithms. It helps researchers and practitioners
evaluate the generalization capability of machine learning models beyond digit recognition
tasks.

Applications:
Fashion-MNIST is often used to benchmark and compare the performance of different
machine learning and deep learning models, especially in the context of image
classification and computer vision tasks.
COMPARISON:
Single Layer Network:
Accuracy: 87.3%

Precision: 88%

Recall: 87%

Comments: The Single Layer Network performed reasonably well on the MNIST Fashion
dataset but showed a drop in performance compared to its performance on the digit
recognition dataset.

Multi-Layer Perceptron (MLP):


Accuracy: 87.8%

Precision: 88%

Recall: 88%

Comments: The MLP demonstrated improved performance over the Single Layer Network
on the MNIST Fashion dataset, indicating its ability to capture more complex features
inherent in fashion images.

Fully Connected DNN:


Accuracy: 87.4%

Precision: 88%

Recall: 88%

Comments: The Fully Connected DNN achieved similar performance to the single layer
network on the MNIST Fashion dataset, highlighting its effectiveness in learning
representations from complex data.

You might also like