Workshop in
Computer Science: Smart Agents & Physiological Monitoring
סדנה במדעי המחשב:
סוכנים
חכמים
במערכות מידע
ניטור
פיזיולוגי
0368-3500-40, Autumn
2014-15
Tuesday 13-15, Schreiber 7
Prof. Nathan Intrator & Mr. Eddie Aronovich
Contact info
Nathan Intrator nin@post.tau.ac.il 03
640 7598
Eddie Aronovich eddiea@cs.tau.ac.il
03 640 6915
Outline for next
three weeks
10/03/2015 General overview – Eddie
12/03/2015 9-10:30 Personal meetings of groups
at Nathan’s office (Schreiber 221)
17/03/2015 Hadoop and other distributed
processing, cloud tools, text analysis
18/03/2015 12:30-2 Personal meetings of
groups at Nathan’s office – project finalization
Please use the subject: Workshop 2015b
in all your emails
Projects
1.
Read My Dog – Ben Bogin, May Michelson & Matan Eyal
2.
Effect of Brain
stimulation on Attention – Itai Adami,
Yair Levi & Guy Rosenmann
Develop a game which will test the user concentration level. The
game will include a spinning colored arrow, surrounded by a colored ring. The
user objective is to touch the screen each time the arrow points to the
matching color in the ring. Data from
various game parameters will be collected from about 30 subjects where about 20
of them are will perform before and after brain stimulation.
3.
Smart Agent for Users
Review - Ayala Igra & Yoav Porat
The browser plugin scans product's reviews,
summarizes them to keywords and detects fake reviews.
The general idea of the plugin is DIDM
(Data-Informed decision-making).
Target audience: the consumers (us) and the
platforms (Amazon, Ebay).
We use Amazon API and Data Mining & NLP
algorithms ('NLPK' and 'Pattern' Python packages)
4.
Flu Tweet - Or Sayag & Ofir Moskovich
5. Mood filter on facebook personal feeder - Rina & Omar
Muhammad Serh
6.
Native Language
Identification Engine – Lee Levi & Yael Avinum
An API for
companies that have multi-language online sales and support, which identifies
in real-time, from English text, what the native language of the customer is.
Our tentative milestones are as follows:
·
Offline Work (until 10.5) : Finalize the machine learning approach, implement and
test the abilities of the engine.
·
API Implementation (until 7.6):
Create an API for the engine that a company could use on top of their own code.
·
GUI (until 21.6): Implement a
multi-user website which includes API documentation, billing, information, and
potentially, user-specific analytics.
7.
Algorithm based Music
Playlists - Maayan
Shvo & Avishag Banai
Description missing.
Milestones:
a.
Prepare the data-gathering
infrastructure and begin to parse and retrieve data from our various sources
b.
Process and analyze the data using
Map-Reduce technique
c.
Finalize data representation and create
GUI (multi-user website)
8.
EEG reading during Oculus VR stimulation –
Sapir Caduri
9.
Physiological
monitoring for neuromarketing – Reut Ben Mocha, Raz
Lugassi, & Tom Sisso
The project focuses on developing an application to advertising
and marketing agencies, to examine the quality of their product (the
commercials). The application will contain information on several aspects: the
level of concentration during the ad (in sections outline specific time),
points of interest in model groups (age, gender) etc. The output will be based
on the analysis of information received from EEG sensor combine with Eye
Follower, connected to a heterogeneous sample, approximately 30 people, while
watching an ad.
Tools to be
available for physiological monitoring
1.
Two EEG boxes with AWS connection
2.
Two watches from MioAlpha
(intense physical activity)
The workshop will
include two main topics:
·
Smart web agents and big data mining
·
Physiological monitoring
There are several
goals for the workshop:
·
Learn a new topic and be able to
program novel functions often assimilating a number of different modalities.
·
Learn to work on a project from design
to validation, writing documentation and presentation.
·
Learn to work as a team, synchronizing
work with another team member and collaborating.
·
Learn to consider all aspects of a
project, from defining the customers, their needs potential solutions and some
business aspects.
Students should
assume that each such project is a potential seed of a startup and focus on
that in their final presentation.
A key feature of
the workshop is to learn how to present your ideas: What is best to present,
and how to do it in the most concise, clear and persuasive manner.
Therefore, there
will be emphasis on the lectures of week 4-5 and week 12-13 where each group
will be presenting to the rest of the groups. Participations of all groups in
the presentations of the other groups is mandatory.
The grade will
mostly be determined on the quality of these two presentations and the ability
to present them (so you should rehearse and make sure they go smooth). The presentations can be given in Hebrew, but
the slides should be in English.
Each PowerPoint
presentation will be accompanied with a Word document and code.
Code should be
self-explanatory (lots of comments) and in the smallest amount of files
(preferably a single file with multiple functions).
All presentation
material will be uploaded on a web page. One group will take upon themselves to
be in charge of the web page.
Participants of
the workshop will develop applications that utilize
· Web crawlers.
· Social networks
API: Facebook, Google+, Linkedin, Twitter (Youtube, flicker).
· Extraction of
data from web sources such as NASDAQ, various government and research
institutes (some will require development of parser etc).
· Text Analysis:
parsing, sentiment analysis, key words extraction etc.
· Clustering and
indexing methods.
· Fast search
tools.
· Parallel
processing e.g. Hadoop or Condor.
· There may be a
project in data collection and analysis of physiological measurements
· Medical and
life-style monitoring applications
· Psychophysical
experiments and data collection w
· Attention and
understanding monitoring during e-learning
(Of course, not
all groups will deal with all above tools)
A good starting
point for technical reading is the Wikipedia page on Text mining
For some administrative
general ideas, see last
year’s workshop
Time line
· Week 1-3: frontal
lectures, brief review on some tools
· Week 2-3:
splitting into groups and choosing a topic
· Week 4-5: Design
review presentations
· Week 8-9: Mid
project progress
· Week 12-13: Final
Presentations
The projects will
be based on the individual background of each participant and will thus, vary
in nature. Projects will have an emphasis either on programming (Matlab, Java, C, Python, etc) or
will be more emphasizing the data mining aspect.
The projects will
be done mainly in couples and in rare cases in triples. Each of the participants will be assigned a specific
part and in the first presentation, each will present their expected part in
the project.
There is room to
combine a project on building a presentation that is optimized for a cell phone
– Iphone (IOS), Android, Windows-8 etc.
Design Review
·
10 min presentation for each group.
All have to participate
·
Problem description (The problem you
are going to solve)
·
Problem background: Size, severity,
who are the consumers and who is going to pay for your solution (how will you
make money out of your solution)
·
Current solution
·
Your solution (From the view point of
the user and the payer (in case these are different)
o
General Overview on your approach
o
The GUI the user see (data description
and screen information)
o
The data you will use
o
The algorithm
o
The code you need to write – describe
the off-the-shelf code and what you need to do to make this work. Other things you need to do, e.g. collect
data, obtain devices etc.
o
3 Milestones for the next 8 weeks (who
does what and when)
Guidelines for Final Presentation (Total
20 min)
·
General problem description and detailed
problem that you are solving (5min)
·
Specific technique and algorithms used (10min)
·
Working demonstration (5 min)
For Reference,
Last Semester Projects
1.
Effect of brain stimulation on
attention
For Reference,
Last Semester Projects
1.
Gal Arnon & Ynon Flum – Heart rate monitoring using MIO Alpha & Android
a.
Literature: heart rate monitoring, HRV, MIO Alpha, BLE Android
communication & programing
b.
Communicate with mio-alpha (pairing) and
read continuous data, address disconnections
c.
Implement two (or more) algorithms from the literature for HRV
(including adaptation)
d.
GUI for sleep and for workout and alerts via sms
and email to designated users
2.
Ram Kalendarev & Sergei + Sergey Panchenko -- SMI
eye tracker – facial (and other objects) familiarity: recognition vs. confusion
a.
Sergie Perlstein – connecting
the BT and analyzing the eeg, meet him for the code
to run familiar faces
b.
Sergey Panchenko - Build the gui
c.
Literature: EEG in general, P300, face detection, EYE Tracker, BT,
d.
Code for connecting the BT EEG to a PC
e.
Code for connecting the EYE tracker to a PC
f.
Based on the eye tracker, detect when a person is looking at a face (or other
object)
g.
At that point analyze the EEG (row
+ Processed) and obtain an output
3.
Ariel Bronner Ron Kirill Nataly
a.
-- Analysis of music (MIDI) tag (cluster) via
accord changes, comparison with human clustering
b.
-- Electronic trading : dynamic combination of technical experts for
better prediction and trend detection.
4.
Yuval Sarna Saimon
Michelson -- Sentiment analysis
on twits for market prediction. Combination
of influential tweets + sentiment analysis. Test forecasting power, compare to
simple predictor based on majority.
5.
Nadine Farah & Hanen Rashed – Investment
Sentiment analysis from News and Company Announcements
Suggested projects
Medical and Life-style - codes will be
in python this year
1. Attention during
computer games
2. Check this movie
about attention while viewing images: Paul Sajda
3. Attention
monitoring during e-learning
4. Mobile EEG
monitor – mobile side
a.
Real-time analysis of the quality of
electrode contact and 50/60Hz noise
b.
Quick reconnection of the BT after
failure (range problem)
c.
Easy acquisition
d.
Data analysis and feature detection
e.
Emphasis on slick GUI and neuro
feedback
f.
Info emanation (sms
etc)
5. Mobile EEG and
Pulse Monitor – Server side
a.
Enable packet sending off information
to a server
b.
Packets can be sent every second, every
minute of every 5 min.
c.
GUI on the server side with feature
alerts
6. EEG + Pulse +
temperature + skin conductance + accelerometer life style monitor
a.
Students can suggest ways to use this
monitor
7.
Code for supporting two BT devices at
the same time
Big Data -
Analytics
8. Analyze
earthquake data
a.
Code in python to replace the Matlab
b.
Code to download off line from
iris.edu
9.
General Sentiment Analysis tool
a.
A general tool receiving a list of
site to extract information from
b.
A list of words to monitor, phrases, sentiment
c.
Produce easy to interpret graphical
representation
d.
Enable clustering of text based on
contents, sentiment
e.
Should be able to provide similar
results to the ones in the article
when given the right data base and keywords
f.
Should be able to extract important words and
associations
10.
Analyze the social protest ("מחאה
חברתית")
messages exchange.
a.
Clicks in top management and boards of
Directors of public Companies.
11.
A
collection of projects around
smart agents, for example an agent that reserves place in a restaurant, based
on your preferences
12.
Financial data: Volatility estimation
between periods. This project requires collecting and analyzing the volatility
and predicting the value of some indexes in the long run and in the short run.
It requires some statistical background.
Sensors
·
BlueTooth + A/D Roving Networks (For the EEG sensor)
·
BlueTooth + A/D KC Wirefree (For
EEG) (old KC-5100) User guide
firmware
·
EMANT380
Bluetooth DAQ (24Bit) for various applications
Examples
·
Heart rate Mio
Alpha
· NBT SMI Eye tracker
Lecture
material
Slides about Unix Tools
Article on Data
Mining (Oded Netzer) TF-IDF
Good design and
development practice
The following
scheme suggests the right order and the necessary steps required to obtain a
good product or functioning unit. The process always starts with understanding
user needs. Too often the wrong problems are solved and the wrong needs are
addressed. The rest is self-explanatory.
Please pay attention to each block in your intermediate and final
reports.
Free code
http://www.rocketdownload.com/freeware.php?q=arrhythmias+ecg