Machine Learning in Intelligent Transportation

Session 1: Introduction

Ahmad Haj Mosa

PwC Austria & Alpen Adria Universität Klagenfurt

Klagenfurt 2020

Contents

• Backgrounds

   

• History of Artificial Intelligence

   

• Definitions

   

• Terminologies

   

• Machine Learning Tasks

   

• Machine Learning Branches

   

• Syllabus and Assessments

   

Objectives of this course

Text

Deep Learning Theory​

Natural Language Processing

DL based Machine Vision

Reinforcement  Learning

Machine  Learning fundamentals

Statistical Machine Learning

Conventional Machine Vision

Course Plan

Text

Linear Regression

Logistics Regression

Shallow/Deep Neural Network

Back propogation algorithms

Convolutional Neural Network

Recurrent  Neural Network

Semantic Segmentation

Generative Adversarial Network

Course Plan

• Day 1:

  • Linear Regression

  • Logistic Regression

  • Shallow Neural Networks

  • Backpropogation Algorithm

  • Variance-Bias Tradeoff

• Day 2:

  • CNN

  • Semantic Segmentation

  • Autoencoders

  • Mask CNN

• Day 3:

  • Recurrent NN

  • LSTMs

  • NLP

  • Transformers

• Day 4:

  • Deep RL

  • AlphaGo Zero

  • GANs

History of Artificial Intelligence 

History of Artificial Intelligence 

AI technologies

Applications

Learning to recognize spoken words:

All of the most successful speech recognition systems employ machine learning in some form.

Alexa

Google Assistant

Zenbo Robot

Applications

Self Driving Cars

Nvidia, Uber, and Tesla all use deep learning as one of the main model in their self driving cars

Applications

Medical Imaging 

Deep Learning is playing a key role in the future of medical imaging dignoses 

Applications

Terminologies

     Data

• Refers to digital (sensor, transactions or text data)


• It could be the output of a device that detects and responds to some type of input from the physical environment .


• It could be a set conversation data collected from a messenger database.

Terminologies

Decision Function/Hypothesis

• A decision function is a function which takes features as input and gives a decision as output


• Decision functions are a mathematical models that maps inputs to targets

In case of self driving cars: the output of the decision function is the steering angle

In case of voice recognition: the output selects a predicted action of a set of actions

Example 3:

• Considering the house price example let us choose a simple linear decision function:

Decision Function/Hypothesis

Terminologies

Model Training/Learning

• In order to learn the decision function \(y(x) \) we require a set of training examples, each example \(i\) contains a specific output value (house price) \(y_i\) and the related features value (house size) \(x_i\) 


• Then the training process estimates the weights \(\omega \) so that the function \(y_i\) fits the data with the minimum possible error

Terminologies

Performance/Cost/Loss

• A measurement of the cost to the performance task of making a prediction \(y^{\prime}\) when the actual label is \(y\) :


• For example one of the most used cost function is the mean squared error:

where \(e_t = ( y^{\prime}- y_t) \)

Machine Learning Tasks

1. Classification

Machine learning task are categorized by the target of the studies problems as follows:

is the problem of identifying to which of a set of categories (sub-populations) a new observation

Machine Learning Tasks

1. Regression

Machine learning task are categorized by the target of the studies problems as follows:

is the problem of identifying the relationship (mathematical model) among different variables

Machine Learning Tasks

3. Clustering

Machine learning task are categorized by the target of the studies problems as follows:

is the task of grouping a set of objects in such a way that objects in the same group (called a cluster) are more similar (in some sense or another) to each other than to those in other groups (clusters)

Machine Learning Tasks

4. Time Series Forecast

Machine learning task are categorized by the target of the studies problems as follows:

Time series forecasting is the use of a model to predict future values based on previously observed values

Machine Learning Branches

The computer is presented with example inputs and their desired outputs, and the goal is to learn a decision function that maps inputs to outputs.

No labels are given to the learning algorithm, leaving it on its own to find structure in its input. Unsupervised learning can be a goal in itself (discovering hidden patterns in data) or a means towards an end (feature learning).

where the teacher gives an incomplete training signal: a training set with some (often many) of the target outputs missing.

Supervised

Unsupervised

      Semi Supervised

Machine Learning Branches

A computer program interacts with a dynamic environment in which it must perform a certain goal (such as driving a vehicle or playing a game against an opponent[4]:3). The program is provided feedback in terms of rewards and punishments as it navigates its problem space

is a branch of machine learning that attempt to model high level abstractions. DL can be supervised, Unsupervised, Semi Supervised and Reinforcement learning. The common model used in DL is Neural Networks

Reinforcement

Deep Learning