We will discuss history and recent developments in the study of the phase structure of noncommutative scalar fields. Apart from the usual disorder and uniform order phases, the theory exhibits a third phase, which survives the commutative limit.
This is connected to the UV/IR mixing of the noncommutative theory. We will rewrite the fuzzy theory as a modified quartic
matrix model, with extra multitrace terms in the action and perform saddle point analysis of the theory. Our goal will be to locate the triple point of the theory and to reconstruct the numerically obtained phase diagram. This goal will be successfully reached at the end of the talk.
This is a slide of My talk at Kyoto Nonclassical Logic Workshop (19, November, 2015). This is based on my paper "A constructive naive set theory and infinity" which was accepted to Notre Dame Journal of Formal Logic.
Xavier Amatriain, VP of Engineering, Quora at MLconf SEA - 5/01/15MLconf
Machine learning applications for growing the world’s knowledge at Quora: At Quora our mission is to “share and grow the world’s knowledge”. We want to do this by getting the right questions to the right people to answer them, but also by getting the existing answers to people who are interested in them. In order to accomplish this we need to build a complex ecosystem where we value issues such as content quality, engagement, demand, interests, or reputation. It is not possible to build a system like this unless most of the process are highly automated and scalable. We are fortunate though to have lots of very good quality data on which to build machine learning solutions that can help address all of the previous requirements.
In this talk I will describe some interesting uses of machine learning at Quora that range from different recommendation approaches such as personalized ranking to classifiers built to detect duplicate questions or spam. I will describe some of the modeling and feature engineering approaches that go into building these systems. I will also share some of the challenges faced when building such a large-scale knowledge base of human-generated knowledge.
We will discuss history and recent developments in the study of the phase structure of noncommutative scalar fields. Apart from the usual disorder and uniform order phases, the theory exhibits a third phase, which survives the commutative limit.
This is connected to the UV/IR mixing of the noncommutative theory. We will rewrite the fuzzy theory as a modified quartic
matrix model, with extra multitrace terms in the action and perform saddle point analysis of the theory. Our goal will be to locate the triple point of the theory and to reconstruct the numerically obtained phase diagram. This goal will be successfully reached at the end of the talk.
This is a slide of My talk at Kyoto Nonclassical Logic Workshop (19, November, 2015). This is based on my paper "A constructive naive set theory and infinity" which was accepted to Notre Dame Journal of Formal Logic.
Xavier Amatriain, VP of Engineering, Quora at MLconf SEA - 5/01/15MLconf
Machine learning applications for growing the world’s knowledge at Quora: At Quora our mission is to “share and grow the world’s knowledge”. We want to do this by getting the right questions to the right people to answer them, but also by getting the existing answers to people who are interested in them. In order to accomplish this we need to build a complex ecosystem where we value issues such as content quality, engagement, demand, interests, or reputation. It is not possible to build a system like this unless most of the process are highly automated and scalable. We are fortunate though to have lots of very good quality data on which to build machine learning solutions that can help address all of the previous requirements.
In this talk I will describe some interesting uses of machine learning at Quora that range from different recommendation approaches such as personalized ranking to classifiers built to detect duplicate questions or spam. I will describe some of the modeling and feature engineering approaches that go into building these systems. I will also share some of the challenges faced when building such a large-scale knowledge base of human-generated knowledge.
Next Generation “Treatment Learning” (finding the diamonds in the dust)CS, NcState
Q: How have dummies (like me) managed to gain (some) control over a (seemingly) complex world?
A:The world is simpler than we think.
◆ Models contain clumps
◆ A few collar variables decide which clumps to use.
Abstract. This manuscript provides a brief introduction to Functional Analysis. It covers basic Hilbert and Banach space theory including Lebesgue spaces and their duals (no knowledge about Lebesgue integration is assumed).
Next Generation “Treatment Learning” (finding the diamonds in the dust)CS, NcState
Q: How have dummies (like me) managed to gain (some) control over a (seemingly) complex world?
A:The world is simpler than we think.
◆ Models contain clumps
◆ A few collar variables decide which clumps to use.
Abstract. This manuscript provides a brief introduction to Functional Analysis. It covers basic Hilbert and Banach space theory including Lebesgue spaces and their duals (no knowledge about Lebesgue integration is assumed).
Presentation at "Emerging problems in particle phenomenology" workshop held at CUNY on April 11, 2010. Has sensitivity of Jets+MET searches for 7 TeV LHC.
This is a parallel presentation from SUSY09 in June 09 on Composite Inelastic Dark Matter. It proposes a model that reconciles various direct detection dark matter experiments.
1. The Post-Model Building Era
and
Simplified Models
Modeling and the LHC
Wuppertal, Germany January 28, 2012
Jay Wacker
SLAC
2. Question:
How should we motivate LHC searches for
signatures of physics beyond the Standard Model?
3. The Challenge Facing the LHC
Last Year
300 Trillion Collisions
1 Billion Recorded Collisions
4. Very hard to make general predictions
Space of experimental signatures is very high
Njets < 12
x
Nleptons < 5 x 3 ( pT, η, φ) ~600 dimensions
x
Nphotons < 4
Sparsely populated
Can’t calculate predictions accurately in this full space of signatures
6. But theories have a high dimensional
parameter space...
MSSM has ~100 parameters
Allowed parameter space has 19 parameters
mSUGRA has 5, but introduces theory prejudice
7. “Theory”
Set of rules based upon principles
used for predicting outcomes
Most Model Building is Theory Building
use principles to create new theories
(naturalness, supersymmetry, unification)
8. “Model”
A representation of a system
Not necessarily physical
Nevents f(Mγγ)
Background
Mγγ
9. Theory vs. Model
Complete vs. Incomplete?
Complete Theory
Answer any physical question
Ultimate goal of theoretical physics
23. Any theory comes with Cutoff
Above Λ, theory may be arbitrarily complicated
Insensitive to Cutoff scale physics at low energies
Is the backdrop for all theory building
Cannot write down complete theory with a straight face
Using hypothetical principles to create new theories
24. Given that we can’t discover the
complete theory of nature,
how do we propose models?
Can parameterize all deviations from
Standard Model
LSM(A, B, ... ) + LNon-Renormalizable (A, B, ... )
25. Given that we can’t discover the
complete theory of nature,
how do we propose models?
Can parameterize all deviations from
Standard Model
LSM(A, B, ... ) + LNon-Renormalizable (A, B, ... )
We usually want to explore
EAB > Mφ
Need to incorporate φ into model
26. Modern Vision of
Theories Beyond the Standard Model
Λ Scale theory is no longer valid
Energy
φ New particles to be discovered
SM What we’ve already seen
Λ can be low ~ 10 TeV Λ can be high ~ 1016 TeV
27. Theory Building In Practice
Pick a problem
Build a theory that solves it
Make predictions for experiment
28. Theory Building In Practice
Pick a problem
Build a theory that solves it
Make predictions for experiment
Argue about which theory is better while waiting
30. The Hierarchy Problem
>50% of motivation for past 35 years
Technicolor Susy
1978
1981
31. The Hierarchy Problem
>50% of motivation for past 35 years
Technicolor Susy
1978
1981
1991
32. The Hierarchy Problem
>50% of motivation for past 35 years
Technicolor Susy Large ED RS Small ED
1978
1981
1991
1998
33. The Hierarchy Problem
>50% of motivation for past 35 years
Technicolor Susy Large ED RS Small ED LH
1978
1981
1991
1998
2002
2012
34. The Hierarchy Problem
>50% of motivation for past 35 years
Technicolor Susy Large ED RS Small ED LH
1978
1981
1991
1998
2002
2012
35. Implications for Experimental Searches
Technicolor Susy Large ED RS Small ED LH
1978
1981
1991
Could enumerate theories
1998 600
500
2002
Hd 2 2 1/2
(µ +m0)
Hu
Mass [GeV]
400
Lots of effort on the specific theories 300
200
M2
M1
M3
m1/2
squarks
100 m0
sleptons
2012 0
2 4 6 8 10 12
Log10(Q/1 GeV)
14 16 18
36. Implications for Experimental Searches
Technicolor Susy Large ED RS Small ED LH
1978
1981
1991
1998
2002
Drowning in Possibilities
2012
37. Belief in any single theory or paradigm
is at all-time low
Ntheories Belief
time
Just examples of possibilities
Model Building Era successful, but over
38. Huge pain for experimentalists
Enormous work to test each theory
Models help motivate where how
separate signal from background
Want to go to the Post-Model Building Era
39. Huge pain for experimentalists
Enormous work to test each theory
Models help motivate where how
separate signal from background
Want to go to the Post-Model Building Era
Is this theory-ladeness acceptable/necessary?
40. Need a way of simplifying theories
Theories Models
41. Simplified Models
Models that are based upon
well-established principles
(e.g. local quantum field theories that contain Standard Model)
Not based upon principles
i.e. there is not explicit physical motivation
Purpose: Reduce Theory-Ladeness
43. Type 1: Narrowly Focused Searches
Theory Space
Signature Space
Experimental
Searches
44. Type 2: Redundant Theories
Theory Space
Signature Space
Experimental
Searches
45. Simplified Models
Start with Standard Model
Postulate relevant particles for a search
Start with 1,2 or 3 new particles
Write down most general theory
Usually small number of parameters
46. Simplified Models
Can capture essential features of existing models
Notice unexplored corners of theory space
from lack of imagination
No burden of top-down motivation
No Principles
47. Simplified Model Example
MASS
˜
g color octet majorana
fermion (“Gluino”) THREE-BODY DECAY
¯
qq
˜
g 0
˜
q 1
˜ neutral majorana
fermion (“LSP”)
52. Gluino Pair Production
j
j
q ¯
q
p g ˜
g
˜
g ET
p g ˜
g
¯
q
q
j
j
Multijets + Missing Energy
53. Common Susy Search Strategy
Base searches on mSUGRA Supersymmetry
mg = 7m
˜ 0
Not general
Risk Type 1 Failure
54. Allowed us to place limits on new theories
with little data
˜
g ¯
qq LHC 70 nb-1
!prod = 3!" NLO-QCD 100 pb
!prod = !" NLO-QCD 200 pb
!prod = 0.3 !" NLO-QCD
300 pb
!prod = 0.1 !" NLO-QCD 500 pb
Sample theory
1 nb
Tevatron
2 nb
mSUGRA
56. Has led to more searches
Modified Triggering
More kinematic regions searched
Unfortunately, no discoveries (yet)
57. Summary of Simplified Models
Too many theories to search for
Simp. Mods.: axes for decomposing all theories
Reduce theory prejudice
Represent natural extension of
Effective Field Theory to the LHC
In the Discovery Era
Construct incomplete models to fit data
When incomplete model doesn’t work extend model
Then construct Theory (understand Principles)