South America Memes: as “Fábricas de Memes” e suas dinâmicas
Xin Li LDCSEE WVU Spring 2009 2015-6-3 From Gene to Meme Example of Memes: -Cultural ideas, symbols...
-
date post
19-Dec-2015 -
Category
Documents
-
view
215 -
download
2
Transcript of Xin Li LDCSEE WVU Spring 2009 2015-6-3 From Gene to Meme Example of Memes: -Cultural ideas, symbols...
Xin Li LDCSEE WVU Spring 2009Xin Li LDCSEE WVU Spring 200923/4/1823/4/18
From Gene to MemeFrom Gene to Meme
Example of Memes:
-Cultural ideas, symbols-Religions-Languages-Mathematics (e.g.,Gaussian distribution)-Scientific understanding (e.g.,Quantum mechanics)-Technological achievements (e.g.,Atomic force microscopy)-Engineering designs (e.g., ipod,Wii, iphone)
Xin Li LDCSEE WVU Spring 2009Xin Li LDCSEE WVU Spring 200923/4/1823/4/18
An Objective Measurement of An Objective Measurement of Meme’s FitnessMeme’s Fitness
Industry Academia
Xin Li LDCSEE WVU Spring 2009Xin Li LDCSEE WVU Spring 200923/4/1823/4/18
What is in Common?
Steve Jobs Herbert Simon
Xin Li LDCSEE WVU Spring 2009Xin Li LDCSEE WVU Spring 200923/4/1823/4/18
Xin Li LDCSEE WVU Spring 2009Xin Li LDCSEE WVU Spring 2009
Theme of This Talk“Life is about connecting dots”– in “Staying Hungry, Staying Foolish” Stanford
Commencement Address by Steve Jobs in 2005
Scientific research is also about connecting dots– Search is an important component part of
scientific research (where are the dots? how are they related?)
– Re-search often reveals hidden relationship among isolated dots that is not known before
Xin Li LDCSEE WVU Spring 2009Xin Li LDCSEE WVU Spring 200923/4/1823/4/18
Image processing
Statisticalphysics
Communication
Networking
CognitiveScience
geometry
PRAM/Microarray
STM/AFM
MRI/PET
Control
analysis
Chemicaloscillation
algebra
statistics
Image Processing as the ShowcaseScience Technology
Engineering
Mathematics
Xin Li LDCSEE WVU Spring 2009Xin Li LDCSEE WVU Spring 200923/4/1823/4/18
Image Processing: at the Intersection Image Processing: at the Intersection of Science, Technology, Engineering of Science, Technology, Engineering
and Mathematics (STEM)and Mathematics (STEM)
+
our starting point
Xin Li LDCSEE WVU Spring 2009Xin Li LDCSEE WVU Spring 2009
So how should you choose your So how should you choose your technical field?technical field?
Outside environment plays some roleOutside environment plays some role– Emerging areas tend to attract more resources Emerging areas tend to attract more resources
than traditional fieldsthan traditional fields– Every dept. has its focused areasEvery dept. has its focused areas
Learn yourself betterLearn yourself better– Good at theory or experiment/application?Good at theory or experiment/application?– Good at algebraic or geometric thinking?Good at algebraic or geometric thinking?– Good at depth-first or width-first reasoning?Good at depth-first or width-first reasoning?
Find a good matchFind a good match23/4/1823/4/18
Xin Li LDCSEE WVU Spring 2009Xin Li LDCSEE WVU Spring 2009
Taking Myself as an Example
Entered Princeton ISS group in 1996 Very little research experience in my undergraduate study (BS thesis is on speech coding)Information theory or signal processing?– Princeton EE is really strong in theory (not to mention
Math and Physics)– Majority of ISS students will take the theory path even
by doing TAs due to limited research funding in the area of information theory
– Graduate students in Princeton EE from India are also really good at theory
Xin Li LDCSEE WVU Spring 2009Xin Li LDCSEE WVU Spring 2009
Information Theory vs. Image Processing
IEEE TIP– H-index=148– 1992-present– Most influential papers:
image watermarking, image coding, image segmentation
IEEE TIT– H-index=214– 1963-present– Most influential papers:
cryptography, space-time codes, error-correcting codes, wavelets, ...
Xin Li LDCSEE WVU Spring 2009Xin Li LDCSEE WVU Spring 2009
Highly-cited Papers related to Highly-cited Papers related to Image ProcessingImage Processing
Markov Random Field (Geman and Markov Random Field (Geman and Geman 1984) >8000 citationsGeman 1984) >8000 citations
Wavelet theory (Daubechies, Mallat, Wavelet theory (Daubechies, Mallat, Vetterli …) >180,000 citationsVetterli …) >180,000 citations
Why do they last?Why do they last?
23/4/1823/4/18
“the most fruitful areas for the growth of sciences were those which had been neglected as a no-man’s land between the various established fields.”
–Norbert Wiener
Xin Li LDCSEE WVU Spring 2009Xin Li LDCSEE WVU Spring 2009
Connection 1: MRFConnection 1: MRFPixels vs. ParticlesPixels vs. Particles
23/4/1823/4/18
Pixel value = 0 or 1 Spin direction = up or down
Xin Li LDCSEE WVU Spring 2009Xin Li LDCSEE WVU Spring 200923/4/1823/4/18
A Little Bit History of Ising A Little Bit History of Ising ModelModel
Proposed by Ising in his PhD thesis in 1925Proposed by Ising in his PhD thesis in 1925
2D Ising model was analytically solved by L. 2D Ising model was analytically solved by L. Onsager in 1944 (who won the Nobel Prize Onsager in 1944 (who won the Nobel Prize in 1968)in 1968)
Phase transition behavior investigated by Phase transition behavior investigated by Yang and Lee in 1950sYang and Lee in 1950s
related to renormalization theory pioneered related to renormalization theory pioneered by RG Wilson (who won the Nobel Prize in by RG Wilson (who won the Nobel Prize in 1982)1982)
Xin Li LDCSEE WVU Spring 2009Xin Li LDCSEE WVU Spring 200923/4/1823/4/18
Apply Ising Model to ImagesApply Ising Model to Images
Applied to image restoration by Geman Applied to image restoration by Geman and Geman in 1984 (bring statistical and Geman in 1984 (bring statistical mechanics to engineering)mechanics to engineering)
Stirred up lots of interestStirred up lots of interest– More powerful image models (line process, More powerful image models (line process,
higher-order MRF)higher-order MRF)– More efficient optimization algorithms (Gibbs More efficient optimization algorithms (Gibbs
sampling, Swendsen-Wang, Wolff algorithm)sampling, Swendsen-Wang, Wolff algorithm)– New applicationsNew applications
Xin Li LDCSEE WVU Spring 2009Xin Li LDCSEE WVU Spring 2009
Image ExampleImage Example
23/4/1823/4/18
original
noisy restored
Monte-CarloOptimization(minimize E)
Xin Li LDCSEE WVU Spring 2009Xin Li LDCSEE WVU Spring 200923/4/1823/4/18
Connection to Hopfield NetworksConnection to Hopfield Networks
Why is this model so influential?
The first-order approximation ofassociative memory in brain theory
Prof. Hopfield gave a talk at WVU on Mar. 13, 2007 titled “How Do We Think So Fast? From Neurons to Brain Computations,”
Xin Li LDCSEE WVU Spring 2009Xin Li LDCSEE WVU Spring 2009
Statistical Mechanics and ITStatistical Mechanics and ITShannon was the first to recognize the connection Shannon was the first to recognize the connection between statistical mechanics and communication theorybetween statistical mechanics and communication theory
Connection with statistical mechanics also exists for Turbo Connection with statistical mechanics also exists for Turbo codes (belief propagation is related to Bethe free energy)codes (belief propagation is related to Bethe free energy)
““Multiuser detection and statistical mechanics” (Guo and Multiuser detection and statistical mechanics” (Guo and Verdu’ 2003)Verdu’ 2003)
““Evolution and structure of the Internet: A statistical Evolution and structure of the Internet: A statistical physics approach” (R Pastor-Satorras and A Vespignani‘ physics approach” (R Pastor-Satorras and A Vespignani‘ 2004)2004)
““Statistical mechanics of complex networks” (R. Albert’ Statistical mechanics of complex networks” (R. Albert’ PhD thesis in 2001)PhD thesis in 2001)
23/4/1823/4/18
Xin Li LDCSEE WVU Spring 2009Xin Li LDCSEE WVU Spring 2009
If you think you have understood If you think you have understood entropyentropy
23/4/1823/4/18
““My greatest concern was what to call it. I thought of calling it My greatest concern was what to call it. I thought of calling it ‘information’, but the word was overly used, so I decided to call it ‘information’, but the word was overly used, so I decided to call it ‘uncertainty’. When I discussed it with John von Neumann, he had ‘uncertainty’. When I discussed it with John von Neumann, he had a better idea. Von Neumann told me, ‘You should call it a better idea. Von Neumann told me, ‘You should call it entropyentropy, , for two reasons. In the first place your uncertainty function has for two reasons. In the first place your uncertainty function has been used in statistical mechanics under that name, so it already been used in statistical mechanics under that name, so it already has a name. has a name. In the second place, and more important, nobody In the second place, and more important, nobody knows what entropy really is, so in a debate you will always have knows what entropy really is, so in a debate you will always have the advantagethe advantage. . ””
-Conversation between Claude Shannon and -Conversation between Claude Shannon and John von NeumannJohn von Neumann regarding what name to give to the “measure of uncertainty” or regarding what name to give to the “measure of uncertainty” or
attenuation in phone-line signals (1949)attenuation in phone-line signals (1949)
Xin Li LDCSEE WVU Spring 2009Xin Li LDCSEE WVU Spring 2009
Connection II: Wavelet Theory Connection II: Wavelet Theory and Image Processingand Image Processing
Wavelet theory was established in late Wavelet theory was established in late 1980s by mathematicians, computer 1980s by mathematicians, computer scientists and electrical engineers togetherscientists and electrical engineers together
The most successful application of The most successful application of wavelets is likely to be lossy image wavelets is likely to be lossy image compression (e.g., JPEG2000)compression (e.g., JPEG2000)– Also popular in other processing tasks such Also popular in other processing tasks such
as segmentation, denoising and retrievalas segmentation, denoising and retrieval
The question is: Why? The question is: Why?
Xin Li LDCSEE WVU Spring 2009Xin Li LDCSEE WVU Spring 2009
Where do Wavelets Come from?Where do Wavelets Come from?Before wavelet, people used Short-Time FT Before wavelet, people used Short-Time FT to analyze transient signalsto analyze transient signals
J. Morlet – a geophysical engineer at a J. Morlet – a geophysical engineer at a French oil company came up with an French oil company came up with an alternative approach which was recognized alternative approach which was recognized by Grossmann – Daubechies’ advisorby Grossmann – Daubechies’ advisor
S. Mallat – a graduate student at Penn met S. Mallat – a graduate student at Penn met Y. Meyer’s student and recognized its Y. Meyer’s student and recognized its connection to multi-resolution analysisconnection to multi-resolution analysis
23/4/1823/4/18
Xin Li LDCSEE WVU Spring 2009Xin Li LDCSEE WVU Spring 2009
Laplacian Pyramids invented by RCA Engineers
Xin Li LDCSEE WVU Spring 2009Xin Li LDCSEE WVU Spring 2009
At the Intersection of Math, CS At the Intersection of Math, CS and EEand EE
Math: construction of basis functions with good Math: construction of basis functions with good localization property in both time and frequencylocalization property in both time and frequency
CS: decomposes images under a multi-resolution CS: decomposes images under a multi-resolution analysis framework in analogy to HVSanalysis framework in analogy to HVS
EE: analysis-and-synthesis filter banks used by TV EE: analysis-and-synthesis filter banks used by TV engineersengineers
Merge of roots: a new tool for data/signal analysisMerge of roots: a new tool for data/signal analysis
Different perspectives: deterministic (Besov-space Different perspectives: deterministic (Besov-space functions) vs. statistical (heavy-tail distributions)functions) vs. statistical (heavy-tail distributions)
23/4/1823/4/18
Xin Li LDCSEE WVU Spring 2009Xin Li LDCSEE WVU Spring 2009
Why Wavelets for Images?Why Wavelets for Images?
23/4/1823/4/18
Math: Besov-space function, statistics: sparse component analysis, neuroscience:Independent components of natural scenes
Xin Li LDCSEE WVU Spring 2009Xin Li LDCSEE WVU Spring 2009
Beyond Image ProcessingBeyond Image ProcessingStatistics: nonparametric regressionStatistics: nonparametric regression
Graphics: progressive mesh compressionGraphics: progressive mesh compression
Turbulence: one of the most complicated Turbulence: one of the most complicated phenomenon in naturephenomenon in nature
Astronomy: hierarchical clustering theory of Astronomy: hierarchical clustering theory of galaxy formationgalaxy formation
Biomedical: MRI, EEG, PET, mammographyBiomedical: MRI, EEG, PET, mammography
Acoustic: computer music analysis Acoustic: computer music analysis
23/4/1823/4/18
Xin Li LDCSEE WVU Spring 2009Xin Li LDCSEE WVU Spring 2009
What is Missing in Wavelet What is Missing in Wavelet Models?Models?
23/4/1823/4/18
DWT
sign flip
IWT
Xin Li LDCSEE WVU Spring 2009Xin Li LDCSEE WVU Spring 2009
Connection III: Complex Connection III: Complex Networks and Image processingNetworks and Image processing
Common assumption made by MRF and Common assumption made by MRF and wavelet models: locality or Markovianwavelet models: locality or Markovian
Most existing physical laws are defined locally; Most existing physical laws are defined locally; but what about but what about nonlocalitynonlocality??
A great mystery in brain science is how it A great mystery in brain science is how it collectively processes local informationcollectively processes local information– Speed of nerve impulse transmission is much Speed of nerve impulse transmission is much
slower than that of logic gatesslower than that of logic gates– The power consumption of neural system is also The power consumption of neural system is also
much more efficient much more efficient
23/4/1823/4/18
Xin Li LDCSEE WVU Spring 2009Xin Li LDCSEE WVU Spring 2009
Networks of NeuronsNetworks of Neurons
23/4/1823/4/18
Xin Li LDCSEE WVU Spring 2009Xin Li LDCSEE WVU Spring 2009
Complex NetworksComplex Networks
Internet Internet
World Wide WebWorld Wide Web
Movie actor collaboration networkMovie actor collaboration network
Science collaboration networkScience collaboration network
Citation networksCitation networks
Cellular networksCellular networks
Ecological networksEcological networks
Power networksPower networks23/4/1823/4/18
Xin Li LDCSEE WVU Spring 2009Xin Li LDCSEE WVU Spring 2009
How is it related to Image How is it related to Image Processing?Processing?
23/4/1823/4/18
B11
B22
B14B13B12
B41
B31
B21
B33B32
B23 B24
B34
B44B43B42
f3f2
f1
Bij
i
j1
2
3
4
1 2 3 4
Parallel and Distributed Processing (PDP)or connectionism was at the foundation of neural networks
Xin Li LDCSEE WVU Spring 2009Xin Li LDCSEE WVU Spring 2009
How is it Different from NN?How is it Different from NN?What role does time play?What role does time play?– Temporal binding hypothesis in neuroscienceTemporal binding hypothesis in neuroscience– Synchronization of nonlinear oscillators in chemical, Synchronization of nonlinear oscillators in chemical,
biological and physical systemsbiological and physical systems
What role does feedback play?What role does feedback play?– As important as feedforwardAs important as feedforward– Mountcastle’s uniformity principle in psychologyMountcastle’s uniformity principle in psychology
Why does the network have to be hierarchical?Why does the network have to be hierarchical?– Natural world is organized in a hierarchical fashionNatural world is organized in a hierarchical fashion– Our perception of natural world is the consequence of mapping Our perception of natural world is the consequence of mapping
from outside (physical stimuli) to inside (synaptic connections)from outside (physical stimuli) to inside (synaptic connections)
23/4/1823/4/18
Xin Li LDCSEE WVU Spring 2009Xin Li LDCSEE WVU Spring 2009
29.06dB 31.56dB 34.96dB
x y DT KR FG1
28.46dB 31.16dB 36.51dB
17.90dB 18.49dB 29.25dB
26.04dB 24.63dB 29.91dB
Experiment 1: Compressed SensingExperiment 1: Compressed Sensing
DT- DelauneyTriangle-based(griddata under MATLAB)
KR- KernalRegression-based(Takeda et al.IEEE TIP 2007w/o parameteroptimization)
1X. Li, “Patch-based image interpolation: algorithms and applications,” Inter. Workshop on Local and Non-Local Approximation (LNLA)’2008
25% kept
Xin Li LDCSEE WVU Spring 2009Xin Li LDCSEE WVU Spring 2009
Experiment 2: Image CodingExperiment 2: Image Coding
JPEG-decoded at rate of 0.32bpp(PSNR=32.07dB)
SFG-enhanced at rate of 0.32bpp(PSNR=33.22dB)
SPIHT-decoded at rate of 0.20bpp(PSNR=26.18dB)
SFG-enhanced at rate of 0.20bpp(PSNR=27.33dB)
Maximum-Likelihood (ML) Decoding
Maximum a Posterior (MAP) Decoding
Xin Li LDCSEE WVU Spring 2009Xin Li LDCSEE WVU Spring 2009
Unfulfilled Connections (I)Unfulfilled Connections (I)
23/4/1823/4/18
Bayer Pattern(US3,971,065)
Cone distribution in human retina
CCD sensor design (engineering) could benefit from the organizationalprinciple of cones in human retina (biology)
Xin Li LDCSEE WVU Spring 2009Xin Li LDCSEE WVU Spring 2009 3434
Unfulfilled Connection (II)Unfulfilled Connection (II)
Q: Can we generate a HDR image (16bpp) by a standard camera?A: Yes, adjust the exposure and fuse multiple LDR images together
Xin Li LDCSEE WVU Spring 2009Xin Li LDCSEE WVU Spring 2009 3535
High Dynamic Range ImagingHigh Dynamic Range Imaging
Note that any commercial display devices we see these days are NOT HDR
Xin Li LDCSEE WVU Spring 2009Xin Li LDCSEE WVU Spring 2009
Unfulfilled Connection (III)Unfulfilled Connection (III)
23/4/1823/4/18
Visual perception might be the first small step towards human intelligencebut it will be a huge leap in human intelligence (can brain understand brain?)
Xin Li LDCSEE WVU Spring 2009Xin Li LDCSEE WVU Spring 200923/4/1823/4/18
Summary and ConclusionsSummary and ConclusionsEvery theory, technology or system has its own Every theory, technology or system has its own evolution pathevolution path– Understand its connection is the most difficult yet important Understand its connection is the most difficult yet important
task task – My learning about image processing is still evolving, but I My learning about image processing is still evolving, but I
am hoping its principle can be also applied to other am hoping its principle can be also applied to other technical field such as communicationtechnical field such as communication
““If you truly believe that God creates this world in a If you truly believe that God creates this world in a unified fashion, when you get stuck with a problem, unified fashion, when you get stuck with a problem, seek your inspiration from around: nature, art and seek your inspiration from around: nature, art and other sciences. Essentially, the principles are the other sciences. Essentially, the principles are the same. ”same. ”
Xin Li LDCSEE WVU Spring 2009Xin Li LDCSEE WVU Spring 2009
Research vs. Development
Good development/programming skills are a plus but secondary to good analytical/logical reasoning skills in my own assessment
Implementation skills should be viewed at the same level as mathematical skills; they are both technical tools but cannot replace scientific vision/understanding
``Knowledge and productivity are like compound interest.'' –Richard Hamming
Xin Li LDCSEE WVU Spring 2009Xin Li LDCSEE WVU Spring 2009
How Good do You Need to at How Good do You Need to at Mathematics?Mathematics?
““Never be overwhelmed by the Never be overwhelmed by the mathematics other people are boasting in mathematics other people are boasting in their papers if you are in engineering: the their papers if you are in engineering: the more equations, the fewer ideas”more equations, the fewer ideas”
Mathematics is a language – you cannot Mathematics is a language – you cannot communicate well if you don’t master it; communicate well if you don’t master it; but you cannot advance science by simply but you cannot advance science by simply playing with mathematics playing with mathematics
23/4/1823/4/18
http://masterxinli.wordpress.com/2008/09/15/how-good-do-you-need-to-be-at-mathematics/
Xin Li LDCSEE WVU Spring 2009Xin Li LDCSEE WVU Spring 2009
Why does engineering/math/science education in the US suck?
http://headrush.typepad.com/creating_passionate_users/2006/11/why_does_engine.html