EFFORT MODELING FOR REINFORCEMENT LEARNING 04/18-ncheema/CV.pdf · SELECTED SE RVICE & VO LUNTEERI NG SELECTED AWA RDS & SCH OLAR SHIP S Full list of publications, awards & service

N O S H A B A - C H E E M A -

[email protected]

CONTACT

Homepagehttps://mpi-inf.mpg.de/~ncheema/

Saarland Uni & Intl. Max-Planck Re-search School for Computer Science04/16-present » Visual Computing M. Sc. » Computer Science M. Sc.

GPA German: 1.7, GPA US: ~3.5

EDUCATION

Ney York University - School of Engineering09/14-12/14 » Foreign semester

GPA US: 3.7, GPA German: ~1.4

University of Appl. Sciences Bremen09/12-06/16 » Media Computer Science B. Sc.

GPA German: 1.3, GPA US: ~3.8

-

Aalto University - School of Science04/18-09/18 » Master-Thesis Student

TECHNICAL SKILLS

C++ ●●●●○ Python ●●●●○ Matlab ●●●●○ C# ●●●●○ Java ●●●●○

Swift ●●●○○ PHP ●●●○○ CUDA ●●●○○ JavaScript ●●○○○ React-Native ●●○○○

Libraries

OpenGL, OpenMP, .NET, Unity

Operating SystemsWindows 7, 8, 10; Ubuntu 14.04 LTS, 16.04 LTS; Mac OS X; iOS; Android

Programming Languages

Beginner ●●●○○ Expert

LANGUAGE SKILLS

(experienced), Punjabi (fundamental)

EFFORT MODELING FOR REINFORCEMENT LEARNINGVisiting Researcher at Aalto University &

» Implementation of Three-Compartment Model by Frey et al. using Monte Carlo in-formed tree search by Rajamäki & Hämäläinen for reinforcement learning.

» Control cost is computed according to current target load and fatigue levels. » Maximum torque allowed at a joint is computed according to current fatigue levels. » Fatigue levels are computed using Three-Compartment Model.

Languages used: C++Frameworks & Libraries used: Open Dynamics Engine

Reference: Prof. Dr. Perttu Hämäläinen - Prof. Dr. Jaakko Lehtinen - [email protected]

SEMANTIC SEGMENTATION OF MOTION CAPTURE DATA

»semantic segmentation of motion capturing data.

» Implementation of motion capture data to RGB image representation. » » Training was done on clean, as well as noisy labels.

Languages used: Python

Reference: Prof. Dr. Philipp Slusallek - [email protected]

NEURO-COMPUTATION IN THE VISUAL PERIPHERY Research Assistant at University of Bremen &

» Implemented new Caffe layer to compute local summary statistics to synthesize foveat-ed versions from white noise of given input images.

» Synthesis was done with a synthesis-by-analysis algorithm using the L-BFGS-B optimizer. The gradient with respect to the input image was computed with the standard for-ward-backward pass of the neural network.

» Determined optimal layer for statistical representations in VGG-networks to use for tex-ture and foveated image synthesis.

Languages used: Python, C++ & CUDAFrameworks & Libraries used: BVLC Caffe

Reference: Dr. Christoph Zetzsche - [email protected]

RAYTRACING-ENGINECourse Project

» Acceleration using BVH structure with Surface Area Heuristic & OpenMP for-loops. » Support of the following materials: mirror material (fuzzy & normal), glass, combined

materials, cook-torrence material, textures, procedural textures. » Support of the following lights: Area light, projective light. » Support of the following cameras: Projective, Orthographic, Fish-Eye, Depth of Field

(using distributed raytracing). » Support of custom 3D rigid body physics engine with distance joints.

Languages used: C++Frameworks & Libraries used: SFML 2.2 & OpenMP

Reference: Prof. Dr. Philipp Slusallek - [email protected]

2D-GAME-ENGINEPersonal Open Source Project

» Acceleration using Quadtree, Spawners, Look-Up-Tables, GPU shaders. » Support of circles, and convex & concave polyhrdrons. » . » Rigid body collision done via seperating axis theorem & voronoi regions. » Support of distance & seesaw joints.

Languages used: C# & GLSLFrameworks & Libraries used: SFML 2.2, .NET & OpenMPRepository: github.com/noshaba/2D-GameEngine

Reference: Thomas Jahn - [email protected]

RELEVENT PROJECTS

04/18-present

11/17-present

09/15-03/17

10/16-02/17

03/15-07/15

University of Bremen - Cognitive Neuroinformatics09/15-03/17 » Student Research Assistant

See: Project - “Neuro-computation in the visual periphery“ & recent papersReference:Dr. Christoph [email protected]

RECENT POSIT IONS

Intelligence Bremen (DFKI GmbH)09/15-12/15 » Intern

See: Project - “Neuro-computation in the visual periphery“ & recent papersReference:Prof. Dr. Udo [email protected]

University of Applied Sciences BremenSummer Term 2014 & 2015 » Applied Mathematics

Supervisor: Prof. Dr. Martin [email protected]

TEACHING

Intelligence Saarbrücken (DFKI GmbH)11/17 - present » Research Assistant

See: Project - “Semantic segmentation of motion capture data“ & recent papersReference:Prof. Dr. Philipp [email protected]

NVIDIA DLI MunichGTC Europe 2017 » Deep Learning classes

Reference: Ana [email protected]

HOBBIES

Painting, Coding, Gaming, 3D Modeling, Cosplaying, Pole Dancing, Cat Videos

Aalto University School of Science04/18 - 09/18 » Visiting Researcher

See: Project - “Effort modeling for reinforcement learning”Reference:Prof. Dr. Perttu Hämäläinen

Prof. Dr. Jaakko [email protected]

University of SaarlandWinter Term 2017 » Basics of Signal Processing

Supervisor: Prof. Dr. Dietrich [email protected]

DILATED TEMPORAL FULLY-CONVOLUTIONAL NEURAL NETWORK FOR SEMANTIC SEGMENTATION OF MOTION CAPTURE DATAACM SIGGRAPH/Eurographics SCA 2018 - Posters (Best Poster Award) » N. Cheema, S. Hosseini, J. Sprenger, E. Herrmann, H. Du, K. Fischer & P. Slusallek

NEURAL COMPUTATION OF STATISTICAL IMAGE PROPERTIES IN PERIPHERAL VISIONMODVIS 2017 » C. Zetzsche, R. Rosenholtz, N. Cheema, K. Gadzicki, L. Fridman & K. Schill

OPTIMUM STATISTICAL REPRESENTATION OBTAINED FROM AN INTERMEDIATE FEATURE LEVEL OF THE VISUAL HIERARCHYPROCEEDINGS OF KOGWIS 2016 » N. Cheema, L. Fridman, R. Rosenholtz & C. Zetzsche

RECENT PUBLICATIONS

ELECTED-STUDENT-COUNCIL-MEMBER 04/17-04/18Compiuter Science Department - Saarland University » Events & PR, elected member from Obtober 2017 to April 2018

Reference: CS Student Council - [email protected]

ESPR17-CAMP-CONSULTANT 02/17-09/17European Summer Program on Rationality 2017 » Organizational member of camp for gifted high-school students

Reference: Jaime Sevilla Molina - [email protected]

SIGGRAPH-STUDENT-VOLUNTEER 07/16-07/16SIGGRAPH 2016 » Volunteer at information booth, papers & VR-village

Reference: Christian Wittorf - [email protected]

NIPS-REVIEWER 06/16-07/16NIPS 2016 » Volunteered to review three papers

Reference: Isabelle Guyon - [email protected]

BEST POSTER PRESENTATION AWARD AT ACM SCA 2018 07/18ACM SIGGRAPH/Eurographics Symposium on Computer Animation 2018

» Best poster presentation award (see papers)Reference: Mélina Skouras - [email protected]

MAX-PLANCK INSTITUTE FOR INFORMATICS FELLOWSHIP 04/16-presentInternational Max-Planck Research School for Computer Science (IMPRS-CS)

» Fellowship from IMPRS-CS at the Max-Planck Institute for InformaticsReference: Jennifer Gerling - [email protected]

DEUTSCHLANDSTIPENDIUM 09/12-02/16University of Applied Sciences Bremen, Telekom Germany & Rotary Club » Scholarship from the German Government & several companies for studies

Reference: Jürgen-Peter Henck - [email protected]

SELECTED SERVICE & VOLUNTEERING

SELECTED AWARDS & SCHOLARSHIPS

Full list of publications, awards & service on https://mpi-inf.mpg.de/~ncheema/

FINE-GRAPHINED SEMANTIC SEGMENTATION OF MOTION CAPTUREDATA USING DILATED TEMPORAL FULLY-CONVOLUTIONAL NETWORKSEurographics 2019 - Short Papers » N. Cheema, S. Hosseini, J. Sprenger, E. Herrmann, H. Du, K. Fischer & P. Slusallek

ADAPTIVE GAUSSIAN MIXTURE TRAJECTORY MODEL FOR PHYSICAL MODEL CONTROL USING MOTION CAPTURE DATAACM SIGGRAPH Symosium on Interactive3D Graphics and Games 2019 » E. Herrmann, H. Du, N. Cheema , J. Sprenger, S. Hosseini, K. Fischer & P. Slusallek