Wednesday 1:55 p.m.–2:35 p.m.

Beating the world's traffic with Machine Learning

Yam Peleg

Audience level:
Intermediate

Description

This talk is about understanding from the ground up the Essentials of Machine Learning Algorithms. The main issue that will be addressed in this talk is predicting traffic jams and in order to minimize the time taken to reach a specific point. in order to do so, the essentials machine learning algorithms will be explained and used live on stage.

Abstract

The main issue that will be addressed in this talk is predicting traffic jams and in order to minimize the time taken to reach a specific point. The talk itself will be humorous and will address a funny problem: "I hate traffic jams! I want to avoid them at all costs!".

The talk will be divided to chapters:

Chapter 1: Introduction - Problem presentation.
Topics:
 - The problem 
 - Big Data in general
 - Machine learning in general
Chapter 2: The data-set.
Topics:
 - Simple traffic data-set will be presented 
 - Data mining
 - Data Quality 
 - Visualizing Datasets
 - Classification VS Regression
Chapter 3: Basic Machine Learning.
Topics:
 - Linear Regression (Demo)
 - Polynomial Regression (Demo)
 - Overfitting 
 - Cross Validation (Demo)
 - Supervised Learning
Chapter 4: Feature Engineering.
Topics:     
-Features VS Raw Data
- Normalization
- Tokenization, Serialization
Chapter 5: Classification Basics.
Topics:
-From Regression to classification 
-Logistic Regression (Demo)
Chapter 6: Feature Learning.
Topics:     
-K-Means
- PCA
- LLE
- ICA
Chapter 7: SVMs
Topics:
-Support Vector Machines (Demo)
-Kernel Based Methods (Demo)
Chapter 8: Hyper parameter Optimization
Topics:
-Grid Search (Demo)
-Random Search
-Gradient Based Optimization
Chapter 9: KNN
Topics:
-K – Nearest Neighbors (Demo)
Chapter10: Random Forests:
Topics:
-Decision Trees (Demo)
-Random Forests (Demo)
-Entropy
Chapter 11: Ensemble learning.
Topics:
        - Bagging
        - Boosting 
        - Bayesian parameter averaging
- Final Live Demo: Beating the traffic jam with an incredibly over-killing algorithm.

Sponsors


Become a sponsor.