Data Science Introduction: NDSC Edition

This course is a specialised version of our Data Science Introduction (Python), designed specifically for the National Data Science Challenge organised by Shopee.

  • TBD
  • 4 sessions
  • 3.5 hours/session
Data Science Introduction: NDSC Edition
$2,000

Price before GST

About this course

About this Course:

This course is a specialised version of our Data Science Introduction (Python), designed specifically for the National Data Science Challenge organized by Shopee—the largest data science competition in Singapore. In this course, students will learn the basic Deep Learning techniques to prepare themselves for the beginner category of the challenge: Product Category Classification.

By the end of the course, students will be able to build simple Deep Learning structures, namely a “classification” or “prediction” engine. Students will also learn to build Deep Learning models with an aim towards product category classification. All coding and projects will be done in class, under the supervision of the instructor.

Whether you are thinking of registering or have registered for the National Data Science Challenge, joining this course will definitely give you the head start you need.

Objectives

- Deep learning basics models and layers
- Linear models and optimisation
- Image classification
- Usage of state-of-the-art deep learning libraries
- Structure of Convolutional Neural Networks (CNN)
- Learning rates and their impact in final results
- CNN’s learning process

Course Plan

1 Linear models and optimization
 1.1 Two dimensional classification
 1.2 Logistic regression
2 Intro to Tensorflow and Keras
 2.1 Keras + tensorboard
 2.2 Tips and tricks
3 Models and layers
1 The problem
 1.1 Image Classification
 1.2 Inputs and Outputs
 1.3 Goal
2 The solution
 2.1 Traditional Computer Vision
 2.2 Deep Learning Computer Vision
3 Deep learning software by name
4 Your First CNN
 4.1 A look at the data
 4.2 Training your first model
5 Convolutional Neural Network
 5.1 Convolutions
 5.2 Convolutional Neural Networks (CNN)
6 Structure of a CNN
 6.1 Convolutional Layer
  6.1.1 Convolutional filters
  6.1.2 Visualization of the Convolutional Layer
  6.1.3 Visualisation of the Receptive Field
 6.2 ReLU (Rectified Linear Units) Layer
  6.2.1 ReLu
 6.3 Pooling Layers
 6.4 Dropout Layer
 6.5 Fully Connected Layer
 6.6 Going further: Convolution Arithmetic
7 CNN in Keras
 7.1 Convolution1D
 7.2 Arguments:
 7.3 Example
 7.4 Convolution2D
  7.4.1 Arguments:
 7.5 Example
 7.6 Dimensions of Conv filters in Keras
  7.6.1 Convolution1D
  7.6.2 Convolution2D
1 Our first CNN
 1.1 Review results
  1.1.1 Most certain cats
  1.1.2 Most certain dogs
 1.2 Freeze layers except last ones
2 Learning
 2.1 The Learning Rate
  2.1.1 Stochastic gradient descent(SGD)
   2.1.1.1 Time-based decay
   2.1.1.2 Momentum
 2.2 RMSprop
 2.3 Adam
 2.4 Adagrad
 2.5 AdaDelta
1 Inside CNN Models
 1.1 Visualizing intermediate activations
 1.2 Visualizing convnet filters
 1.3 Visualizing heatmaps of class activation

Prerequisites

This course requires a basic understanding of Python. Students looking to sign up for this course would have to have graduated from our Python Development Course. A Student Affairs Officer will reach out to you upon registration to confirm your mastery of Python.

There are no open runs for this course at the moment. If you're interested in taking this course, you may join the waitlist and you will be notified when there are vacancies.

UpCode Academy

Attend coding classes taught by true experts working in the industry. Get practical instructions and interact with these practitioners during the classes.

  • 28 Courses
  • 1,159 Students
  • 28 Instructors

Subscribe to our Newsletter