CS401 Transfer Learning PDF
Document Details
Uploaded by SweetNoseFlute
Maynooth University
2024
Ronan Reilly
Tags
Related
Summary
This document provides lecture notes on transfer learning for CS401, a course on machine learning and neural networks. The material covers what transfer learning is, ResNet architecture, fine-tuning for a new task, and image classification. It also includes a link to a dataset used for image classification.
Full Transcript
11/11/2024 CS401: MACHINE LEARNING AND NEURAL NETWORKS TRANSFER LEARNING RONAN REILLY DEPARTMENT OF COMPUTER SCIENCE MAYNOOTH UNIVERSITY [email protected] 1 OVE...
11/11/2024 CS401: MACHINE LEARNING AND NEURAL NETWORKS TRANSFER LEARNING RONAN REILLY DEPARTMENT OF COMPUTER SCIENCE MAYNOOTH UNIVERSITY [email protected] 1 OVERVIEW WHAT IS TRANSFER LEARNING RESNET ARCHITECTURE FINE-TUNING RESNET FOR A NEW TASK CLASSIFYING IMAGES OF CATS AND DOGS A JULIA IMPLEMENTATION 2 1 11/11/2024 TRANSFER LEARNING The reuse of a previously learned model on a new problem is known as transfer learning. It allows the training of deep neural networks with a relatively small amount of data. 3 4 2 11/11/2024 RESNET ResNet-50 is a convolutional neural network that is 50 layers deep. You can load a pre-trained version of the network trained 1.2 million images from the ImageNet database with an average image size of 469x387 pixels. The pretrained network can classify images into 1000 object categories, such as keyboard, mouse, pencil, and many animals. 5 FINE-TUNING RESNET TO CATEGORISE CAT AND DOG IMAGES https://www.kaggle.com/c/dogs-vs-cats/data 6 3 11/11/2024 Go to notebook… 7 4
