TensorFlow (8 op)

Tavoitteet

This course has developed as a practical approach to machine learning and deep learning using TensorFlow. TensorFlow is an open-source software library created by the Google Brain team to make the computing load easier and faster for machine learning and deep learning applications. This course brings students hands-on experience building his/her machine learning models, state-of-the-art image classifiers, and deep neural networks. Furthermore, he/she will learn advanced techniques and algorithms to work with large real-world datasets to prepare for future job opportunities.

This course is 100% virtual thanks to the comprehensive interactive material and content prepared for this course.

The student will pass this course after submitting the required quiz, assignments, and the final project.

Sisältö

1. Introduction:
What is TensorFlow? – TensorFlow 2.x vs. TensorFlow 1.x – Setting up TensorFlow – Getting Started with Google Colab

2. Python Overview
Variables and Operators – Data Structures – Loops and Conditional Statements

3. Building Models using Keras:
What is Keras? – Machine Learning Models in General – Simple Neural Network Sequential Model – Fitting, Evaluation, and Prediction – Simple Computer Vision Model using Neural Networks (Handwriting Recognition) – Tensors vs. Variables – Callbacks

4. Convolutional Neural Networks in TensorFlow:
Why CNN? – CNN Layers – Implementing CNN in TensorFlow – Training and Evaluating CNN for MNIST Dataset

5. Handling Overfitting in TensorFlow:
What is Overfitting? – Regularization: Basics – L1 and L2 Regularizations – Early Stopping – Dropout – Batch Normalization

6. Transfer Learning and TensorFlow Hub:
Saving Models and Weights – Loading Weights – What is Transfer learning? – TensorFlow Hub

7. Final Project

Työmuodot

This course is 100% virtual thanks to the comprehensive interactive material and content prepared for this course.

Course includes:
- Tutorial Videos
- Exercises
- Quiz
- Project
- Self-study

Aika ja paikka

Course is online in TechClass environment and it can be done in own pace.

Oppimateriaali ja suositeltava kirjallisuus

Online.

Opintojaksototeutuksen valinnaiset suoritustavat

N/A

Harjoittelu- ja työelämäyhteistyö

N/A

Tenttien ajankohdat ja uusintamahdollisuudet

Online.

Kansainväliset yhteydet

N/A

Opiskelijan ajankäyttö ja kuormitus

Lectures = 80h
Assignments = 50h
Self-study = 80h
Quiz = 10h
Project = 40h
Essay = 10h
Total = 270 hours

Sisällön jaksotus

Up to Student her-/himself.

Arviointimenetelmät ja arvioinnin perusteet

Exercises 30%
Quizzes 20%
Project 40%
Essay 10%

Arviointiasteikko

Hyväksytty/Hylätty

Arviointikriteerit arvosanalle 1 tyydyttävä

- The student is familiar with TensorFlow’s features for machine/deep learning applications.
- The student knows about the first and the second generations of TensorFlow.
- The student knows how to set up and get started with TensorFlow in the Google Colab environment.
- The student is familiar with the basic syntax of Python and knows how to write simple scripts.
- The student is familiar with the general framework of Keras.
- The student is familiar with machine learning models and their basic concepts.

Arviointikriteerit arvosanalle 3 hyvä

- The student knows how to train simple machine learning models, evaluate them, and make predictions based on them in TensorFlow.
- The student knows how to implement simple neural networks in TensorFlow.
- The student knows the concept of tensors and how they are different from variables.
- The student is familiar with the intuition behind callbacks.
- The student is familiar with the general framework of convolutional neural networks (CNN).
- The student is familiar with different layers of CNN.
- The student is familiar with the concepts of overfitting and regularization.
- The student knows how to implement CNN in TensorFlow for computer vision tasks.

Arviointikriteerit arvosanalle 5 kiitettävä

- The student knows how to analyze the performance of CNN after training.
- The student is familiar with L1 and L2 regularizations and can employ them to avoid overfitting.
- The student understands the concept of early stopping to avoid overfitting.
- The student understands dropout and batch normalization techniques to avoid overfitting.
- The student knows how to transfer learning models work.
- The student is familiar with TensorFlow Hub.