Problem of the week

TensorFlow Tutorial

Python Tutorial

Numpy Tutorial

 

MACHINE LEARNING CONFERENCE

Saturday, April 7, 2018

Hosted by Eastern Michigan University

Time and location: 8:30am - 3:30pm in room 216, Pray Harrold building



Conference Theme
The conference theme for this year is Deep Learning, Theoretical and Practical Aspects.

Conference Goal: To bring together faculty and students, to offer the opportunity of showing their work to others, and to invite them to discussions and prospective future cooperation.

Conference Fee
There is no conference fee this year. The Department of Mathematics&Statistics covers all the costs involving the room and breakfast.

Conference Chair: Ovidiu Calin, Department of Mathematics&Statistics, Eastern Michigan University. If you have any questions please contact the conference chair.


Conference Program:

8:30 - 9:00
Registration and Breakfast, location: entrance hall of Pray Harrold

--------------------------------------------------------------------------------------------------------------------------------------

9:00 - 9:15
Ovidiu Calin,
faculty, EMU: A welcome message and a brief introduction to Machine Learning
Abstract: This is a short message for the students who would like to pursue a career in a field related to machine learning and data science. A few covered topics are: Why the need of machine learning? Why is ML that special? The future of AI.
Slides here


9:20 - 9:55

Khairul Islam and Xinyuan Zhang, faculty and student, EMU: Determinants of Heavy Drinking: Application of Machine Learning in Logistic Regression
Abstract: This study investigates determinants of heavy drinking using a machine learning random forest method and logistic regression analysis. The process involves selecting a set of factors relating to heavy drinking by a random forest approach. Finally, a logistic regression model determines what extent these factors influence heavy drinking.
Slides here


10:00 - 10:30

Tim Burtell, student, EMU: GUI for Designing and Training a Feed Forward Network
Abstract:   I will demonstrate software I have written for creating the structure of a FFN and training the network using a training set input from a file. The FFN uses different activation functions and its arhitecture is adjustable.  
Slides here


10:30 - 11:00

Aaron Bolton and Praveen T W Hettige, students, EMU: Classifying Mosquito Sounds Using Machine Learning Techniques
Abstract: Anopheles mosquitoes are the primary vector of malaria. We use Support Vector Machines, K-Nearest Neighbors and Random Forest to classify 5 mosquito species by features extracted from their wing-beat sounds. Our accuracy is roughly 91% for SVM, 87% for KNN and 89% for Random Forest.


11:00 - 11:35

Ovidiu Calin, faculty, EMU: Some Theoretical and Practical Aspects of Deep Learning
Abstract:
I will talk about how the information flow propagates through the layers of a neural network. The lost information and compresion are described in terms of the mutual information. Some other topics such as network capacity and information bottleneck will be discussed. An application to the MNIST data classification will be included.
Slides here


11:40 - 12:15
Henry Han, faculty, Fordham University:
Implied Volatility Pricing via Integrative Learning
Abstract: With the surge of massive data in finance, implied volatility pricing remains a challenge for its essential role in trading, though few model-driven methods are available in the literature. In this work, we proposed a data driven implied volatility analytics by inventing a novel integrative learning approach. The proposed method integrated different machine learning models to price impolied volatility for various in-the-money options by leveraging the availability of a large amount of data in the market. The proposed approach not only demonstrates its superiority in prediction accuracy, but also a strong model independence by overcoming the generalization issue of traditional model-driven approaches.

12:15 - 13:15
Lunch break
 

13:15 - 13:55

Muhammad Sohaib Arif, student, EMU: YOLO for impact detection in Ice Hockey.
Abstract: YOLO is short for You Only Look Once. This is method that divides an image into a grid and retrieves bounding boxes for each grid using convolution and then regression. The box is comprised of 5 items that are coordinates of the center in x and y, relative height and width of the box and category of the object detected.
I used this method with a dataset that I developed from 3 videos and timestamps that were taken from sensors and then corroborated manually by someone else at EMU.
Slides here

13:55 - 14:20

Syed Hussain, student, EMU:
Abstract: I will present a talk on YOLO (You Only Look Once) object detection model on the LISA Traffic Sign Dataset with an enhancement of the original Data set using a neural network on the Data set which will be for certain weather conditions or for night time. This modification will double the size of the Data set and will be able to detect the Traffic Signs for different weather conditions such as Snow or Rain. This is a joint project with Sohaib Arif also from the Computer Science Department.


14:20 - 14:50

Elijah Nichols, student, EMU: The creation of a feedforward deep neural network to accurately predict the selling price of a used car.
Abstract: I will talk about the creation of a feedforward deep neural network which predicts the selling price of a used car. The model will be trained with various market and product factors such as model year, mileage, color, number and severity of crashes, etc. gathered from Kelley Blue Book and other, similar services for aggregating used car data.
Slides here


14:50 - 15:30

Andrew Ross, faculty, EMU: Ethics, Public Policy, and Machine Learning
Abstract: I will discuss many issues in the ethical use of machine learning, and how it interacts with public policy. Examples include automated parole recommendations and car insurance pricing. We will also talk about general characteristics of contexts that can use ethical/public policy issues in machine learning, and common traps like data that is biases based on existing societal influences. This talk is inspired and based on the book "Weapons of Math Destruction" by Cathy O'Neill and other books.


--------------------------------------------------------------------------------------------------------------------------------------
See you again next year!