A Survey on Analog Models of Computation
Amaury Pouly Joint work with Olivier Bournez
Université de Paris, IRIF, CNRS, F-75013 Paris, France
30 june 2020
Survey: https://arxiv.org/abs/1805.05729
1 / 24 Nowadays: “analog” = continuous/opposite of digital
orthogonal concepts ) even continuous/discrete unclear: hybrid exists )
The meaning of “analog”
Historically: “analog” = by analogy, i.e. same evolution
R L
z b k V (t) q C
m
F(t) F = mz¨ + bz˙ + kz V = Lq¨ + Rq˙ + 1 q , C
2 / 24 The meaning of “analog”
Historically: “analog” = by analogy, i.e. same evolution
R L
z b k V (t) q C
m
F(t) F = mz¨ + bz˙ + kz V = Lq¨ + Rq˙ + 1 q , C
Nowadays: “analog” = continuous/opposite of digital
orthogonal concepts ) even continuous/discrete unclear: hybrid exists )
2 / 24 Some analog machines
Slide Rule
Difference Engine
Antikythera mechanism Linear Planimeter
3 / 24 Some analog machines
Admiralty Fire Control Table ENIAC
Differential Analyzer Kelvin’s Tide Predicter
4 / 24 MathematicalComputabilityNot general purpose model model Analog Circuits Differential Analyzer
Planimetery 0 = f (y) AntikytheraContinuousGPAC y 0 Tide= p(y Predicter) DynamicalAFCT System Difference Engine Discrete Turingyn+1 machine= f (yn) Dynamical System ENIAC Commodore Digital Circuits Slide Rule laptop server supercomputer
Classifying machines/models
time continuous
space
discrete
5 / 24 MathematicalComputabilityNot general purpose model model Analog Circuits Differential Analyzer
Planimetery 0 = f (y) AntikytheraContinuousGPAC y 0 Tide= p(y Predicter) DynamicalAFCT System Difference Engine Discrete Turingyn+1 machine= f (yn) Dynamical System
Slide Rule
Classifying machines/models
time continuous
space
ENIAC Commodore Digital Circuits laptop server discrete supercomputer
5 / 24 MathematicalComputabilityNot general purpose model model
y 0 = f (y) ContinuousGPAC y 0 = p(y) Dynamical System Difference Engine Discrete Turingyn+1 machine= f (yn) Dynamical System
Slide Rule
Classifying machines/models
time continuous Analog Circuits Differential Analyzer Planimeter Antikythera Tide Predicter AFCT
space
ENIAC Commodore Digital Circuits laptop server discrete supercomputer
5 / 24 MathematicalComputabilityNot general purpose model model
y 0 = f (y) ContinuousGPAC y 0 = p(y) Dynamical System
Discrete Turingyn+1 machine= f (yn) Dynamical System
Classifying machines/models
time continuous Analog Circuits Differential Analyzer Planimeter Antikythera Tide Predicter AFCT Difference Engine space
ENIAC Commodore Digital Circuits Slide Rule laptop server discrete supercomputer
5 / 24 MathematicalComputability model model
y 0 = f (y) ContinuousGPAC y 0 = p(y) Dynamical System
Discrete Turingyn+1 machine= f (yn) Dynamical System
Classifying machines/models
time Not general purpose continuous Analog Circuits Differential Analyzer Planimeter Antikythera Tide Predicter AFCT Difference Engine space
ENIAC Commodore Digital Circuits Slide Rule laptop server discrete supercomputer
5 / 24 MathematicalComputabilityNot general purpose model model
y 0 = f (y) ContinuousGPAC y 0 = p(y) Dynamical System
Discrete Turingyn+1 machine= f (yn) Dynamical System
Classifying machines/models
time continuous Analog Circuits Differential Analyzer
space
ENIAC Commodore Digital Circuits laptop server discrete supercomputer
5 / 24 ComputabilityNot general purpose model
GPAC y 0 = p(y)
Turing machine
Classifying machines/models
time continuous Mathematical model Analog Circuits Differential Analyzer = ( ) Continuous y 0 f y Dynamical System
space Discrete yn+1 = f (yn) Dynamical System ENIAC Commodore Digital Circuits laptop server discrete supercomputer
5 / 24 MathematicalNot general purpose model
= ( ) Continuous y 0 f y Dynamical System
Discrete yn+1 = f (yn) Dynamical System
Classifying machines/models
time continuous Computability model Analog Circuits Differential Analyzer
GPAC y 0 = p(y)
space Turing machine
ENIAC Commodore Digital Circuits laptop server discrete supercomputer
5 / 24 R recursive functions Large population protocols Hybrid Systems Timed automata Physarum computing Boolean difference equation models Shannon’s GPAC Black hole models Hopfield’s neural networks Reaction-Diffusion Systems Chemical reaction networks
Continuous Automata Deep learning models Neural networks Blum Shub Smale machines Hybrid systems Natural computing influence dynamics Signal machines
The many many models
time continuous
space Population protocols Finite state automata Turing machines Petri nets Lambda calculus Recursive functions discrete Post systems Cellular automata Chemical reaction networks
6 / 24 R recursive functions Large population protocols Hybrid Systems Timed automata Physarum computing Boolean difference equation models Shannon’s GPAC Black hole models Hopfield’s neural networks Reaction-Diffusion Systems Chemical reaction networks
The many many models
time continuous
space Population protocols Continuous Automata Finite state automata Deep learning models Turing machines Neural networks Petri nets Lambda calculus Blum Shub Smale machines Recursive functions Hybrid systems discrete Post systems Natural computing influence dynamics Cellular automata Signal machines Chemical reaction networks
6 / 24 Boolean difference equation models
The many many models
time R recursive functions continuous Large population protocols Hybrid Systems Timed automata Physarum computing Shannon’s GPAC Black hole models Hopfield’s neural networks Reaction-Diffusion Systems Chemical reaction networks space Population protocols Continuous Automata Finite state automata Deep learning models Turing machines Neural networks Petri nets Lambda calculus Blum Shub Smale machines Recursive functions Hybrid systems discrete Post systems Natural computing influence dynamics Cellular automata Signal machines Chemical reaction networks
6 / 24 The many many models
time R recursive functions continuous Large population protocols Hybrid Systems Timed automata Physarum computing Boolean difference equation models Shannon’s GPAC Black hole models Hopfield’s neural networks Reaction-Diffusion Systems Chemical reaction networks space Population protocols Continuous Automata Finite state automata Deep learning models Turing machines Neural networks Petri nets Lambda calculus Blum Shub Smale machines Recursive functions Hybrid systems discrete Post systems Natural computing influence dynamics Cellular automata Signal machines Chemical reaction networks
6 / 24 ?
quantum analog
Making sense of all these models
logic boolean circuits discrete
recursive Turing lambda functions machines calculus
continuous
“Church” thesis All discrete models are Turing machine-computable.
7 / 24 ?
analog
Making sense of all these models
logic boolean circuits discrete
recursive Turing lambda functions machines calculus
continuous quantum
“Church” thesis All discrete models are Turing machine-computable.
7 / 24 Making sense of all these models
logic boolean circuits discrete
recursive Turing lambda functions machines calculus
?
continuous quantum analog
“Church” thesis ? All models are Turing machine-computable.
Clearly wrong: a single real number (⌦ of Chaitin) is super-Turing pow- erful.
7 / 24 Making sense of all these models
logic boolean circuits discrete
recursive Turing lambda functions machines calculus
?
continuous quantum analog
“Church” thesis ? All physical machine-based models are Turing machine-computable.
Several issues with that statement.
7 / 24 abstraction mathematical model
? proof
physical interpretation computability truth results
I mathematical model = abstraction of a system I properties of model = properties of system 6 I conclusion might be quantitatively or qualitatively wrong
Machine vs mathematical model
physical machine
8 / 24 ? proof
physical interpretation computability truth results
I properties of model = properties of system 6 I conclusion might be quantitatively or qualitatively wrong
Machine vs mathematical model
physical abstraction mathematical machine model
I mathematical model = abstraction of a system
8 / 24 ?
physical interpretation truth
I conclusion might be quantitatively or qualitatively wrong
Machine vs mathematical model
physical abstraction mathematical machine model
proof
computability results
I mathematical model = abstraction of a system I properties of model = properties of system 6
8 / 24 Machine vs mathematical model
physical abstraction mathematical machine model
? proof
physical interpretation computability truth results
I mathematical model = abstraction of a system I properties of model = properties of system 6 I conclusion might be quantitatively or qualitatively wrong
8 / 24 Informal theorem If slowly rotating Kerr black holes exists, one can check consistency of ZFC or solve the Turing halting problem in finite time.
I conclusion: hypercomputations are possible ? Common occurrence in analog models: non-computable reals, Zeno phenomena, ...
Black hole model and hypercomputations
I machine: the universe I model: general relativity
9 / 24 Common occurrence in analog models: non-computable reals, Zeno phenomena, ...
Black hole model and hypercomputations
I machine: the universe I model: general relativity
Informal theorem If slowly rotating Kerr black holes exists, one can check consistency of ZFC or solve the Turing halting problem in finite time.
I conclusion: hypercomputations are possible ?
9 / 24 Black hole model and hypercomputations
I machine: the universe I model: general relativity
Informal theorem If slowly rotating Kerr black holes exists, one can check consistency of ZFC or solve the Turing halting problem in finite time.
I conclusion: hypercomputations are possible ? Common occurrence in analog models: non-computable reals, Zeno phenomena, ... 9 / 24 Physical Church Turing thesis/Thesis M Whatever can be calculated by a machine (with finite data/instructions) is Turing machine-computable.
machine that conforms to the physical laws ) Alternative thesis All reasonable models of computations are equivalent to Turing machines.
Back to the Church thesis
Distinguish machines from models: Actual Church thesis Every effective computation can be carried out by a Turing machine, and vice versa. effective = systematic method in logic/mathematics/CS )
10 / 24 Alternative thesis All reasonable models of computations are equivalent to Turing machines.
Back to the Church thesis
Distinguish machines from models: Actual Church thesis Every effective computation can be carried out by a Turing machine, and vice versa. effective = systematic method in logic/mathematics/CS ) Physical Church Turing thesis/Thesis M Whatever can be calculated by a machine (with finite data/instructions) is Turing machine-computable.
machine that conforms to the physical laws )
10 / 24 Back to the Church thesis
Distinguish machines from models: Actual Church thesis Every effective computation can be carried out by a Turing machine, and vice versa. effective = systematic method in logic/mathematics/CS ) Physical Church Turing thesis/Thesis M Whatever can be calculated by a machine (with finite data/instructions) is Turing machine-computable.
machine that conforms to the physical laws ) Alternative thesis All reasonable models of computations are equivalent to Turing machines.
10 / 24 Observation A system/machine can have several models, all useful, depending on the level of abstraction.
Semantics (assuming law of mass action): I discrete I differential I stochastic
Chemical Reaction Networks (CRNs)
A reaction system is a finite set of I molecular species y1,...,yn f I reactions of the form ai yi bi yi (ai , bi N, f = rate) i ! i 2 Example (any resemblanceP to chemistryP is purely coincidental): 2H + O H O ! 2 C + O CO 2 ! 2
11 / 24 Observation A system/machine can have several models, all useful, depending on the level of abstraction.
Chemical Reaction Networks (CRNs)
A reaction system is a finite set of I molecular species y1,...,yn f I reactions of the form ai yi bi yi (ai , bi N, f = rate) i ! i 2 Example (any resemblanceP to chemistryP is purely coincidental): 2H + O H O ! 2 C + O CO 2 ! 2 Semantics (assuming law of mass action): I discrete I differential I stochastic
11 / 24 Observation A system/machine can have several models, all useful, depending on the level of abstraction.
Chemical Reaction Networks (CRNs)
A reaction system is a finite set of I molecular species y1,...,yn f I reactions of the form ai yi bi yi (ai , bi N, f = rate) i ! i 2 Example (any resemblanceP to chemistryP is purely coincidental): 2H + O H O ! 2 C + O CO 2 ! 2 Semantics (assuming law of mass action): I discrete ! yi = molecule count I differential close to population protocols I stochastic
11 / 24 Observation A system/machine can have several models, all useful, depending on the level of abstraction.
Chemical Reaction Networks (CRNs)
A reaction system is a finite set of I molecular species y1,...,yn f I reactions of the form ai yi bi yi (ai , bi N, f = rate) i ! i 2 Example (any resemblanceP to chemistryP is purely coincidental): 2H + O H O ! 2 C + O CO 2 ! 2 Semantics (assuming law of mass action): discrete I yi = concentration I differential ! polynomial ODEs I stochastic
11 / 24 Observation A system/machine can have several models, all useful, depending on the level of abstraction.
Chemical Reaction Networks (CRNs)
A reaction system is a finite set of I molecular species y1,...,yn f I reactions of the form ai yi bi yi (ai , bi N, f = rate) i ! i 2 Example (any resemblanceP to chemistryP is purely coincidental): 2H + O H O ! 2 C + O CO 2 ! 2 Semantics (assuming law of mass action): I discrete yi = probability distribution I differential stochastic ODEs I stochastic !
11 / 24 Chemical Reaction Networks (CRNs)
A reaction system is a finite set of I molecular species y1,...,yn f I reactions of the form ai yi bi yi (ai , bi N, f = rate) i ! i 2 Example (any resemblanceP to chemistryP is purely coincidental): 2H + O H O ! 2 C + O CO 2 ! 2 Semantics (assuming law of mass action): I discrete I differential I stochastic Observation A system/machine can have several models, all useful, depending on the level of abstraction.
11 / 24 Something is wrong, change the model.
Let’s study it! Even if it exists, Especially if it it cannot be verified. doesn’t exist.
Possible conclusion All reasonable models of computations are equivalent to Turing machines. Hypercomputability results can help us correct models.
Let’s do something useful with it!
Are analog systems capable of hypercomputations? Examples: Black holes, signal machines, hybrid systems
12 / 24 Something is wrong, change the model.
Let’s study it! Even if it exists, Especially if it it cannot be verified. doesn’t exist.
Possible conclusion All reasonable models of computations are equivalent to Turing machines. Hypercomputability results can help us correct models.
Are analog systems capable of hypercomputations? Examples: Black holes, signal machines, hybrid systems
Let’s do something useful with it!
12 / 24 Let’s study it! Even if it exists, Especially if it it cannot be verified. doesn’t exist.
Possible conclusion All reasonable models of computations are equivalent to Turing machines. Hypercomputability results can help us correct models.
Are analog systems capable of hypercomputations? Examples: Black holes, signal machines, hybrid systems
Let’s do something Something is wrong, useful with it! change the model.
12 / 24 Even if it exists, it cannot be verified.
Possible conclusion All reasonable models of computations are equivalent to Turing machines. Hypercomputability results can help us correct models.
Are analog systems capable of hypercomputations? Examples: Black holes, signal machines, hybrid systems
Let’s do something Something is wrong, useful with it! change the model.
Let’s study it! Especially if it doesn’t exist.
12 / 24 Possible conclusion All reasonable models of computations are equivalent to Turing machines. Hypercomputability results can help us correct models.
Are analog systems capable of hypercomputations? Examples: Black holes, signal machines, hybrid systems
Let’s do something Something is wrong, useful with it! change the model.
Let’s study it! Even if it exists, Especially if it it cannot be verified. doesn’t exist.
12 / 24 Are analog systems capable of hypercomputations? Examples: Black holes, signal machines, hybrid systems
Let’s do something Something is wrong, useful with it! change the model.
Let’s study it! Even if it exists, Especially if it it cannot be verified. doesn’t exist.
Possible conclusion All reasonable models of computations are equivalent to Turing machines. Hypercomputability results can help us correct models.
12 / 24 u k k uv v ⇥
u u+v u u v + R R General Purpose Analog Computer, Shannon 1936
y1(t) y(0)=y0, y 0(t)=p(y(t)) t Polynomial Differential Equation, Graça 2004
General Purpose Analog Computer (GPAC)
Differential analyzer
13 / 24 y1(t) y(0)=y0, y 0(t)=p(y(t)) t Polynomial Differential Equation, Graça 2004
General Purpose Analog Computer (GPAC)
u k k uv v ⇥
u u+v u u v + R R General Purpose Analog Computer, Shannon 1936 Differential analyzer
13 / 24 General Purpose Analog Computer (GPAC)
u k k uv v ⇥
u u+v u u v + R R General Purpose Analog Computer, Shannon 1936 Differential analyzer
y1(t) y(0)=y0, y 0(t)=p(y(t)) t Polynomial Differential Equation, Graça 2004
13 / 24 General Purpose Analog Computer (GPAC)
u k k uv v ⇥
u u+v u u v + R R General Purpose Analog Computer, Shannon 1936 Differential analyzer
y1(t) y(0)=y0, y 0(t)=p(y(t)) t Polynomial Differential Equation, Graça 2004
13 / 24 y2 y1 ⇥ R R
g y3 y4 ` ⇥ R 1 ⇥ ⇥ R
y10 = y2 y1 = ✓ y = g y y = ✓˙ 8 20 l 3 8 2 y 0 = y y , y =sin(✓) > 3 2 4 > 3