Description:
Most of the data collected these days are unstructured data and mainly in image format. To make sense of this data special techniques using Big Data and Artificial Intelligence. With the rise of facial recognition use cases for security, surveillance etc., Image recognition has become critical to make interactions between human and machine. Applied Deep Learning for Image processing and Forecasting gives them a practical level of experience, achieved through a combination of about 50% lecture, 50% demo work with student’s participation.
Duration:
3 Days
Course Code: BDT7
Learning Objectives:
After this course, you will be able to:
- Compare AI vs ML vs DL
- Understand TensorFlow and Keras
- Discuss how to identify which kinds of technique to be applied for specific use case
- Understand CNN and RNN techniques
- Understand Convolutional Neural Networks
- Do image recognition tasks
- Understand usage of tools through a AI Demo and hands-on labs.
This course is designed for Software Architects, Developers, Data Engineer, Analyst and Machine Learning Engineer.
- Basic knowledge of command line Linux editors (VI / nano)
- Basic understanding of Machine Learning
- Python experience as applied to the Data Science/Machine Learning space
Course Outline:
- History and Background of AI and ML
- Compare AI vs ML vs DL
- Introduction to neural networks
- The math behind neural networks
- Back propagation
- Understanding the intuition behind neural networks
Introducing TensorFlow
- TensorFlow intro
- TensorFlow Features
- TensorFlow Versions
- GPU and TPU scalability
- Lab: Setting up and Running TensorFlow
- The Tensor: The Basic Unit of TensorFlow
- Introducing Tensors
- TensorFlow Execution Model
- Lab: Learning about Tensors
Introducing Perceptrons
- Single Layer Linear Perceptron Classifier With TensorFlow
- Linear Separability and Xor Problem
- Activation Functions
- Softmax output
- Backpropagation, loss functions, and Gradient Descent
- Lab: Single-Layer Perceptron in TensorFlow
Hidden Layers: Intro to Deep Learning
- Hidden Layers as a solution to XOR problem
- Distributed Training with TensorFlow
- Vanishing Gradient Problem and ReLU
- Loss Functions
- Lab: Feedforward Neural Network Classifier in TensorFlow
High-level Tensorflow: tf.learn
- Using high-level TensorFlow
- Developing a model with tf.learn
- Lab: Developing a tf.learn model
Convolutional Neural Networks in Tensorflow
- Introducing CNNs
- CNNs in Tensorflow
- Lab: CNN apps
Introducing Keras
- What is Keras?
- Using Keras with a Tensorflow Backend
- Lab: Example with a Keras
From Deep Neural Networks to Deep Learning
- Understanding unstructured data
- Image recognition
- Introduction to Convolutional Neural Networks (CNN)
- Convolutional layers
- Pooling layers
- Fully-connected layers
- Hands-on: Using TensorFlow to create a CNN
- Hands-on: Image recognition project
Image processing elements
- Convolutions
- Pooling
- Edge Detection
- De-noising
Time series processing and forecasting elements
- Traditional Time Series forecasting with ARIMA models
- Defining Autocorrelation
- Understanding the Dickey-Fuller Test
Forecasting with TensorFlow and Keras
- Using RNN and LSTM in time series prediction.
- Validation and metrics of Time Series Prediction models
References and Next steps
Structured Activity/Exercises/Case Studies:
- TensorFlow Hands-on
- Keras Hands-on
- Using TensorFlow to create an CNN
- Image Recognition project
Training material provided: Yes (Digital format)
