Deep Learning on GPU Mattias Fält Dept. of Automatic Control Lund Institute of Technology Mattias Fält Deep Learning on GPU Overview What is the difference between CPU and GPU? What is CUDA, and how does it relate to cuBLAS and cuDNN? How is this connected to Deep Learning and Tensorflow? How do I run tensorflow on the GPU? What is a TPU? Mattias Fält Deep Learning on GPU GPU vs CPU http://allegro

https://www.control.lth.se/fileadmin/control/Education/DoctorateProgram/DeepLearning/2016/presentation.pdf - 2025-02-14

https://www.control.lth.se/fileadmin/control/Education/DoctorateProgram/HistoryOfControl/2016/AndronovOnVyshnegradski.pdf - 2025-02-14

Untitled 1 History of Cont rol - Int roduc tion Karl Johan Åström Department of Automatic Control LTH Lund University Int roduc tion 1. Introduction 2. Practical Information 3. A Thumbnail History 4. The Power of Feedback 5. Summary Theme: Those who ignore history are doomed to repeat it. Those who cannot remember the past are condemned to repeat it. (George Santayana) Why bot her? ◮ Fun ◮ Useful

https://www.control.lth.se/fileadmin/control/Education/DoctorateProgram/HistoryOfControl/2016/L01introductioneight.pdf - 2025-02-14

Untitled 1 Process Cont rol Karl Johan Åström Department of Automatic Control LTH Lund University Process Cont rol K. J. Åström 1. Introduction 2. The Industrial Scene 3. Pneumatics 4. Theory? 5. Tuning 6. More Recent Development 7. Summary Theme: Measurement Control Instrumentation and Communication (pneumatic). Lectures 1940 1960 2000 1 Introduction 2 Governors | | | 3 Process Control | | | 4 Ae

https://www.control.lth.se/fileadmin/control/Education/DoctorateProgram/HistoryOfControl/2016/L03ProcessControleight.pdf - 2025-02-14

L05ASMENyquistLecture.pdf A S M E N y q u is t L e c tu re 2 0 0 5 N yq u is t an d H is S em in al P ap er s K ar l J o h an Å st rö m D ep ar tm en t o f M ec h an ic al E n g in ee ri n g U n iv er si ty o f C al if o rn ia S an ta B ar b ar a A S M E N y q u is t L e c tu re 2 0 0 5 H ar ry N yq u is t 18 89 -1 97 6 A G if te d S ci en ti st a n d E n g in ee r Jo h n so n -N yq u is t n o is

https://www.control.lth.se/fileadmin/control/Education/DoctorateProgram/HistoryOfControl/2016/L05ASMENyquistLecturesix.pdf - 2025-02-14

() Automatic Control Emerges Karl Johan Åström Department of Automatic Control LTH Lund University Karl Johan Åström Automatic Control Emerges Automatic Control Emerges K. J. Åström 1 Introduction 2 The Computing Bottleneck 3 State of the Art around 1940 4 WWII 5 Servomechanisms 6 Summary Theme: Unification, theory, and analog computing. Karl Johan Åström Automatic Control Emerges Lectures 1940 19

https://www.control.lth.se/fileadmin/control/Education/DoctorateProgram/HistoryOfControl/2016/L07ControlEmerges.pdf - 2025-02-14

Untitled 1 Automatic Cont rol in Sweden Karl Johan Åström Department of Automatic Control, LTH Lund University Lectures 1940 1960 2000 1 Introduction 2 Governors | | | 3 Process Control | | | 4 Feedback Amplifiers | | | 5 Harry Nyquist | | | 6 Aerospace | | | 7 Automatic Control Emerges ← | | 8 The Second Phase ← ← | 9 Automatic Control in Sweden | | | 10 Automatic Control in Lund | | 11 The Futur

https://www.control.lth.se/fileadmin/control/Education/DoctorateProgram/HistoryOfControl/2016/L09Swedeneight.pdf - 2025-02-14

L11Future.pdf Some personal reflections The Future of Control K. J. Åström Department of Automatic Control LTH Lund University LTH April 24 2012 NAE, AFOSR, IEEE, IFAC LTH April 24 2012 The Systems Perspective In the past steady increases in knowledge has spawned new microdisciplines within engineering. However, contemporary challenges – from biomedical devices to complex manufacturing designs to

https://www.control.lth.se/fileadmin/control/Education/DoctorateProgram/HistoryOfControl/2016/L11Future_8.pdf - 2025-02-14

https://www.control.lth.se/fileadmin/control/Education/DoctorateProgram/HistoryOfControl/2016/M_C_History.pdf - 2025-02-14

governors.dvi ON GOVERNORS J.C. MAXWELL From the Proceedings of the Royal Society, No.100, 1868. A GOVERNOR is a part of a machine by means of which the velocity of the machine is kept nearly uniform, notwithstanding variations in the driving-power or the resistance. Most governors depend on the centrifugal force of a piece connected with a shaft of the machine. When the velocity increases, this f

https://www.control.lth.se/fileadmin/control/Education/DoctorateProgram/HistoryOfControl/2016/MaxwellOnGovernors.pdf - 2025-02-14

https://www.control.lth.se/fileadmin/control/Education/DoctorateProgram/HistoryOfControl/2016/ProfosOnStodola.pdf - 2025-02-14

History of Real Time Systems History of Real Time Systems Gautham Department of Automatic Control, Lund University 1/14 Gautham: History of Real Time Systems Overview Introduction 1940s 1950s 1960s RTOS A look at RTSS Cloud. The future? 2/14 Gautham: History of Real Time Systems Real Time Systems I Real Time Systems describes hardware and software systems subject to a ”real-time constraint”, for e

https://www.control.lth.se/fileadmin/control/Education/DoctorateProgram/HistoryOfControl/2016/hoc_Gautham.pdf - 2025-02-14

1 Lecture 1: Introduction (Karl Johan) 2 Lecture 2: Calculus of variation (CoV) and the Maximum principle In this lecture, we are going to learn the maximum principle. The MP is a type of CoV, so we will first study the classical theory of CoV. Then we will try to move from the classical CoV theory to the optimal control setting, there we will immediately encounter some essential difficulties that

https://www.control.lth.se/fileadmin/control/Education/DoctorateProgram/Optimal_Control/2023/Lec2.pdf - 2025-02-14

4 Lecture 4. Misc topics on MP and the proof of the MP In the previous lecture, we studied the maximum principle. The main result is the following theorem. Theorem 1. Consider the system ẋ = f(x, u) with cost function J(u) = φ(x(tf )) + ∫ tf 0 L(x, u)dt and boundary constraint x(tf ) ∈ M ⊆ Rn Assume f , φ and L are C1 in x. Let (x∗(·), u∗(·)) correspond to the optimal solution to the minimiza- ti

https://www.control.lth.se/fileadmin/control/Education/DoctorateProgram/Optimal_Control/2023/Lecture4.pdf - 2025-02-14

Optimal Control Introduction Karl Johan Åström Department of Automatic Control LTH Lund University Optimal Control K. J. Åström 1. Introduction 2. Calculus of Variations 3. Optimal Control 4. Computations 5. Stochastic Optimal Control 6. Conclusions Theme: Subspecialities A Brief History Early beginning: Bernoulli, Newton, Euler, Lagrange The Golden Era 1930-39: Department of Mathematics at Univer

https://www.control.lth.se/fileadmin/control/Education/DoctorateProgram/Optimal_Control/2023/OptimalControlIntroeight.pdf - 2025-02-14

Exercise for Optimal control – Week 1 Choose two problems to solve. Exercise 1 (Fundamental lemma of CoV). Let f be a real valued function defined on open interval (a, b) and f satisfies ∫ b a f(x)h(x)dx = 0 for all h ∈ Cc(a, b), i.e., h is continuous on (a, b) and its support, i.e., the closure of {x : h(x) ̸= 0} is contained in (a, b). 1) Show that f is identically zero if f is continuous. If f

https://www.control.lth.se/fileadmin/control/Education/DoctorateProgram/Optimal_Control/2023/ex1.pdf - 2025-02-14

Exercise for Optimal control – Week 2 Choose one problem to solve. Disclaimer This is not a complete solution manual. For some of the exercises, we provide only partial answers, especially those involving numerical problems. If one is willing to use the solution manual, one should judge whether the solutions are correct or wrong by him/herself. Exercise 1 (Insect control). Let w(t) and r(t) denote

https://www.control.lth.se/fileadmin/control/Education/DoctorateProgram/Optimal_Control/2023/ex2-sol.pdf - 2025-02-14

Exercise for Optimal control – Week 3 Choose 1.5 problems to solve. Exercise 1. Consider a harmonic oscillator ẍ + x = u whose control is constrained in the interval [−1, 1]. Find an optimal controller u which drives the system at initial state (x(0), ẋ(0)) = (X1, X2) to the origin in minimal time. Draw the phase plot. Exercise 2. Consider a rocket, modeled as a particle of constant mass m movin

https://www.control.lth.se/fileadmin/control/Education/DoctorateProgram/Optimal_Control/2023/ex3.pdf - 2025-02-14

Exercise for Optimal control – Week 4 Choose one problem to solve. Disclaimer This is not a complete solution manual. For some of the exercises, we provide only partial answers, especially those involving numerical problems. If one is willing to use the solution manual, one should judge whether the solutions are correct or wrong by him/herself. Exercise 1. Use tent method to derive the KKT conditi

https://www.control.lth.se/fileadmin/control/Education/DoctorateProgram/Optimal_Control/2023/ex4-sol.pdf - 2025-02-14

Exercise for Optimal control – Week 5 Choose 2 problems to solve. Exercise 1. A public company has in year k profits amounting to xk SEK. The management then distributes uk to the shareholders and invests xk − uk in the company itself. Each SEK invested in such way will increase the company profit by θ > 0 the following year so that xk+1 = xk + θ(xk − uk). Suppose x0 ≥ 0 and 0 ≤ uk ≤ xk so that xk

https://www.control.lth.se/fileadmin/control/Education/DoctorateProgram/Optimal_Control/2023/ex5.pdf - 2025-02-14