You are on page 1of 12

LeNet & AlexNet

LeNet, developed by Yann LeCun et al. in 1989.

The LeNet-5 CNN architecture has seven layers. Three convolutional layers,
two subsampling layers, and two fully linked layers make up the layer
composition.

A 32x32 grayscale image serves as the input for LeNet-5


LeNet - 5
AlexNet:
AlexNet, developed by Alex, Ilya, and Geoffrey Hinton in 2012

AlexNet comes out as a leading architecture.

It comprises eight layers, including five convolutional layers followed by max-


pooling layers, and three fully connected layers.
AlexNet:
AlexNet:
Fully connected layers:
Activation Function:

LeNet primarily uses the sigmoid and tanh activation function in its layers.

AlexNet uses the rectified linear unit (ReLU) activation function.

Using the ReLU as an activation function accelerated the speed of the training
process by almost six times.
Training Method:

LeNet was trained using traditional optimization techniques such as stochastic


gradient descent (SGD).

AlexNet was trained using a variant of SGD called stochastic gradient descent
with momentum (SGDM).

It also employes the regularization technique to avoid overfitting.


Computational Resources:

Due to its relatively shallow architecture, LeNet is less computationally


demanding compared to AlexNet and is suitable for training on less powerful
hardware.

AlexNet's deeper architecture and larger number of parameters require more


computational resources for training and inference, making it more suitable for
high-performance computing environments.
Impact and Recognition:

LeNet was one of the pioneering CNN architectures.


It laid the groundwork for future developments in the field of deep learning.
It was primarily used for handwritten digit recognition tasks.

AlexNet was a breakthrough in the field of computer vision, winning the


ImageNet Large Scale Visual Recognition Challenge (ILSVRC) in 2012 by a
significant margin.
Its success popularized deep learning and paved the way for the development of
more complex CNN architectures.
Thanks <3
Ramadan Mubarak

You might also like