The Internet belongs to everyone. Let’s keep it that way.

Protect Net Neutrality
Loading presentation...

Present Remotely

Send the link below via email or IM

Copy

Present to your audience

Start remote presentation

  • Invited audience members will follow you as you navigate and present
  • People invited to a presentation do not need a Prezi account
  • This link expires 10 minutes after you close the presentation
  • A maximum of 30 users can follow your presentation
  • Learn more about this feature in our knowledge base article

Do you really want to delete this prezi?

Neither you, nor the coeditors you shared it with will be able to recover it again.

DeleteCancel

CoCo and CogLunch

No description
by

Maria Blondet

on 12 November 2015

Comments (0)

Please log in to add your comment.

Report abuse

Transcript of CoCo and CogLunch

B
B
A
A
Conclusions
Highly Accurate with only one electrode
All algorithms extracted robust features
Stable across time
Future Directions
Biometrics
Requirements
Universality
Distinctiveness
Permanence
Collectability
Practical Issues
Decision time
Circumvention
Acceptability
Brainwaves as a Biometric Characteristic
Why?
Why Brainwaves as authentication method?
Safety for the user, not only for the system
What would we need?
Classification
Methods
Behind DIVA
Autoencoder
Principal Component Analysis
Brainwaves
Acquisition

In the search of Distinctiveness...
Details here:
Experiment Design
Raw data Time and Frequency Domains
30Hz LPF Time and Frequency Domains
30Hz HPF Time and Frequency Domains
30 Subjects

70 trials of:
BAG
FISH
MOG
BPW
DVD
TNT
BPW
PGN
Available dataset
Words
Non Words
Acronyms
Illegal Strings
Shown twice
Permanence?
A subset of
15
subjects came back within 1 week and 1 month
later
to repeat the experiment
A simple approach
Neural Network Approach
The Benchmark
Real ERP
Best
sERP
Worst
sERP
Models incremental learning process based on the Rescorla-Wagner (1972) Equilibrium Equations
Fast Neural Network
Cross-correlation
Divergent Autoencoder
Naive Discriminant Learning
Support Vector Machine
Examples
Why is this relevant for Cognitive Science?
Quantify the uniqueness of our cognitive selves
Brainprint: Identifying Unique Features of Neural Activity with Machine Learning
Maria RuizBlondet - Negin Khalifian - Blair C. Armstrong - Zhanpeng Jin - Kenneth J. Kurtz - Sarah Laszlo

Channel: Right Middle Occipital
Output
Hidden
Input
550
200
550
550
...
User 1
User 30
Kurtz, 2007
Kernel: RBF Sigma: 1000
Input
550
30
Output
1 iteration training thanks to Danks equilibrium equations
Comparison with Current Work
98% Accuracy
Oscillatory signals in Gamma band
Use of 60 electrodes
New Machine Learning Algorithms
Explore Permanence
Biometric Specific Protocol
Questions?
This work was supported by awards to S.L. from NSF CAREER-1252975, to Z.J. and S.L. from NSF TWC SBE, the Binghamton University Interdisciplinary Collaborative Grants program, and the Binghamton University Health Sciences Transdisciplinary Area of Excellence, and to B.C.A. from the European Science Commission MC IIF-627784.
Thank You!
Contact: mruizbl1@binghamton.edu
Practical Solution to Duress
ACCESS DENIED
EEG
Commercial Applications
MRI/fMRI
MEG
NIR
Electrical Signal
Hemodynamic Signal
Electrical Signal
Hemodynamic and
Electrical Signal
EEG
Commercial Applications
Electrical Signal
Marta Kutas and Kara D. Federmeier
Time for Meaning: Electrophysiology Provides Insights into the Dynamics of Representation and Processing in Semantic Memory
Kara D. Federmeier, Sarah Laszlo
x 100 Tokens
Per subject
Laszlo & Federmeier, 2011
N400 and Semantic Memory
1st presentation:
Training Dataset
2nd presentation:
Testing Dataset
Trained for 1000 iterations
2 Class Scenario
Subject 1
Subject 1
Subject 1
...
100 tokens
Subject 2
Subject 2
Subject 30
...
3 tokens
x subject
Subject 2
Subject 3
Subject 3
Subject 3
Subject 30
Subject 30
Authorized User
Impostor
Palaniappan & Mandic, 2007
Davies Bouldin index
Brainwaves Acquisition
Classification Methods
Results
50 Trials Averaged
Clown
Bee
Shrimp
Cilantro
Universality
Distinctiveness
Permanence
Collectability
Decision time
Circumvention
Acceptability
Collectability
97% Accuracy
ERPs including N400 component
Use of
ONE
electrode
Current Work
Our Study
Universality
Distinctiveness
Up to 97%
Permanence
Kept up to 89%
Collectability
1 Electrode
Distinctiveness
Permanence
Post-Processing
find subspace with
largest variance
Acceptability
+
Meta
Classifier
Naive
Discriminant
Learning
Divergent
Autoencoder
Cross
Correlation
Voting
Scheme
Final
Output
Analysis of the Data
Subject 21
Subject 21
Accuracy per Subject
Normal Subject (18)
Future Directions
Survey Results (n=7)
Loved
Hated
2
2
2
2
2
3
4
2
Celebrities
1
2
Food
Loved
Hated
Develop a new experiment with subliminal stimuli.
... In the search of bimodal stimuli
System
The Task
Enrollment
Subject 1
Subject 30
...
Identification
Subject 30
?
Subject 1
Verification Task
Real ERP
Best
sERP
Worst
sERP
Full transcript