http://como.vub.ac.be/ 2012-05-17T01:07:35+02:00 http://como.vub.ac.be/ http://como.vub.ac.be/lib/images/favicon.ico text/html 2009-12-14T09:27:21+02:00 http://como.vub.ac.be/teaching:adaptive_systems?rev=1260779241&do=diff Teacher Professor Bernard Manderick Goal To introduce pattern classification and neural networks together with their applications. The student has to be able to: * understand the basic ideas behind these techniques * implement these techniques * apply these techniques to simple problems * evaluate their performance text/html 2011-09-26T16:23:12+02:00 http://como.vub.ac.be/teaching:adaptive_systems_seminar?rev=1317046992&do=diff Teacher Professor Bernard Manderick Time and Place This course takes place in the 2nd semester. Goal To acquire the necessary skills to understand and to make a synthesis of state of the art topics in machine learning and complex dynamical systems. text/html 2012-02-08T13:17:56+02:00 http://como.vub.ac.be/teaching:ai?rev=1328703476&do=diff Teacher Professor: Bernard Manderick Teaching Assistant: David Catteeuw Goal Introduction to Artificial Intelligence. Prerequisites None. Course Material Artificial Intelligence: a modern approach, third edition, S. Russell & P. Norvig. Chapters 1-9, 13. text/html 2008-10-09T08:40:38+02:00 http://como.vub.ac.be/teaching:brugcursus?rev=1223534438&do=diff De taak staat [hier]. Die moet je volgende donderdag in het begin van de les aan mij afgeven. Voor zij die daar niet zullen zijn, drop het ten laatste tegen dan in de postbus Wiskunde in het Centrum voor Studiebegeleiding. Groeten! Ruben text/html 2009-10-06T17:05:45+02:00 http://como.vub.ac.be/teaching:clustering_toolbox?rev=1254841545&do=diff Context Many applications involve finding patterns and relations in massive amounts of data. Examples are fraud detection in bank transactions, customer profiling, processing clinical trial data, etc. One of the “basic” tasks in this domain (viz. Data Mining) is discovering subgroups of similar data. There's several names for this, a.o. clustering, subgroup discovery, automatic classification, etc. text/html 2011-09-27T11:20:29+02:00 http://como.vub.ac.be/teaching:complex_systems?rev=1317115229&do=diff Teacher Professor Bernard Manderick Time and Place There are no lectures. Goal To introduce the students to the mathematical background of the complex dynamics paradigm in artificial intelligence. Prerequisites BA computer science or BA mathematics option computer science or equivalent. Linear algebra and calculus are mathematical requirements. text/html 2009-10-09T17:04:42+02:00 http://como.vub.ac.be/teaching:computation?rev=1255100682&do=diff Teachers Professor: Bernard Manderick (Computation Theory) and Leo van Biesen (Information Theory) Teaching Assistant: David Catteeuw (Computation Theory) Goal (computation theory part) To introduce the basic concepts of the theory of computation in particular the theory of (undecidable) problems and the several complexity classes and to show the relevance of this theory for the computer science practice. text/html 2011-04-26T11:17:45+02:00 http://como.vub.ac.be/teaching:computational_biology?rev=1303809465&do=diff Context Proteins have been compared to neural networks. Like these weighted networks they are claimed of being able to propagate and store information. Their role is to ensure under a variety of circumstances the proper functional behavior of nature's robots, i.e. the cells. text/html 2011-09-29T11:42:48+02:00 http://como.vub.ac.be/teaching:decisioneng?rev=1317289368&do=diff Course code: WE-DINF-13864 Teacher Professors: Philippe Vincke and Yves de Smet Time and Place 2nd semester course. See detailed schedule here. Course description See here. text/html 2011-02-15T17:39:49+02:00 http://como.vub.ac.be/teaching:distributed?rev=1297787989&do=diff text/html 2008-06-04T18:46:05+02:00 http://como.vub.ac.be/teaching:evol_coop?rev=1212597965&do=diff Often the selfish and strong are believed to be favored by natural selection, even though cooperative and fair interactions thrive at all levels of organization in living systems. This project tackles this paradox in the context of Evolutionary Game Theory (EGT), having kin-selection, direct and indirect reciprocity as conceptual starting points. Contrary to what is usual, models will also take into account the intricate ties of modern social networks and its topological evolution. text/html 2011-04-08T15:03:00+02:00 http://como.vub.ac.be/teaching:evolutionary_dynamics?rev=1302267780&do=diff Context Often, the selfish and strong are believed to be favored by natural selection, even though cooperative and fair interactions thrive at all levels of organization in living systems. This project tackles this paradox in the context of Evolutionary Game Theory (EGT), having kin-selection, direct and indirect reciprocity as conceptual starting points. Contrary to what is usual, models will also take into account the intricate ties of modern social networks and its topological evolution. text/html 2010-10-04T15:02:26+02:00 http://como.vub.ac.be/teaching:genome3ba?rev=1286197346&do=diff text/html 2011-09-19T09:57:55+02:00 http://como.vub.ac.be/teaching:grondslagen2?rev=1316419075&do=diff Docent professor Ann Nowé Leerdoelen en Voorkennis Basiskennis theoretische grondslagen van programmeertalen (syntax, berekenbaarheid) overdragen. De student moet inzicht hebben in de mogelijkheden en de beperkingen van reguliere en context vrije talen, en in de corresponderende automaten voor deze talen. text/html 2010-03-16T13:16:55+02:00 http://como.vub.ac.be/teaching:grondslagen2oef?rev=1268741815&do=diff Opgaven en Oplossingen Zie de Pointcarrépagina van het vak voor opgaven en oplossingen Taken De taken dienen om een beeld te geven van wat zou kunnen gevraagd worden op het examen. Ze zijn NIET verplicht. Het is mogelijk om de asssistenten te vragen om ze te verbeteren, maar er worden GEEN oplossingen gegeven. De opgaven voor de taken zullen verschijnen op pointcarré. text/html 2012-02-17T16:43:39+02:00 http://como.vub.ac.be/teaching:heu_opt?rev=1329493419&do=diff Course code: INFOH413 Teacher Professor Thomas Stuetzle Time and Place The course will start on Monday, Feb 13, 2012. The course takes place Wednesdays from 08:00 to 10:00 in P.Forum.H and from 10:00 to 12:00 in different rooms, according to the schedule here. text/html 2010-10-04T15:55:16+02:00 http://como.vub.ac.be/teaching:ipad3ba?rev=1286200516&do=diff text/html 2011-07-27T18:32:22+02:00 http://como.vub.ac.be/teaching:lp?rev=1311784342&do=diff Course lecturer Dr. Coen De Roover Assistant Yann-Michaël De Hauwere Theory <http://soft.vub.ac.be/~cderoove/teaching.html> Book P. Flach, Simply Logical. Intelligent Reasoning by Example, J. Wiley & Sons, 1994, ISBN 0-471-94152-2, available at <http://www.cs.bris.ac.uk/~flach/SimplyLogical.html> Additional exercises and information: text/html 2010-02-16T16:22:59+02:00 http://como.vub.ac.be/teaching:machine_learning?rev=1266333779&do=diff Teacher Professor: Ann Nowé Goal Knowledge and insight The student is acquainted with a range of basic learning algorithms. The student is capable of applying evaluation techniques to estimate the performance of the obtained model and to calculate the performance of an algorithm given a concrete application context using computational learning theory. text/html 2012-01-12T15:09:13+02:00 http://como.vub.ac.be/teaching:mals?rev=1326377353&do=diff Teachers Professor Ann Nowé and Professor Tom Lenaerts Teaching Assistants: Mihail Mihaylov, Peter Vrancx Course together with the ULB. Information available here Time and Place First Semester: Thursdays at 13:00 Place: 10F720 Content The course addresses two general areas of research : individual-based learning and social learning in populations. text/html 2012-02-17T09:32:49+02:00 http://como.vub.ac.be/teaching:mals_seminar?rev=1329467569&do=diff Teacher Professor Ann Nowé Assistant Peter Vrancx Assistant Tim Brys Time and Place Tuesday, 13h CoMo Lab (10G711) Schedule Tuesday 27/09/2011: Short introduction to the course Bi-weekly meetings Tuesday 13-16h. Content This year's course will deal with Robotic Soccer. Robocup is an international competition aimed at developing autonomous robot soccer players. Teams from across the globe compete in a number of different leagues (humanoid, small scale, simulation,...) in order to devel… text/html 2008-05-30T14:32:55+02:00 http://como.vub.ac.be/teaching:marl_glue?rev=1212150775&do=diff Introduction A lot of research within the Computational Modeling Lab is focused on Reinforcement Learning (RL). This machine learning paradigm aims at learning a mapping between situations and actions as to maximize a numerical reward (See T.Mitchell, Machine Learning, 1997). The standard RL model is shown in the figure below. text/html 2012-05-15T12:54:22+02:00 http://como.vub.ac.be/teaching:mathesisproposals?rev=1337079262&do=diff * Reinforcement Learning in Linear-Quadratic Differential Games. * An Android Framework for distributed Reinforcement Learning. * Learning Trajectory Control for Autonomous Vehicles * Using Recurrent Neural Networks for Mobile Robot Navigation * Learning Control for Quadrotor Helicopters * Optimization Algorithm Portfolios * Multi-agent Transfer Learning * Population based Reinforcement Learning * Adaptive Heuristics * Computational Biology proposals * Machine Learnin… text/html 2009-05-08T13:15:18+02:00 http://como.vub.ac.be/teaching:negotiation_scm?rev=1241781318&do=diff Context Supply Chain Management is the process of organising the flow of goods (raw materials, work-in-process, finished products) between locations (suppliers, warehouses, production plants, distribution centres, customers, etc.). Many enterprises outsource this activity to specialised logistics service providers that take complete responsibility over the transportation needs of their customers. Because of the growing complexity of supply chains, a new type of organisation has emerged: Fourt… text/html 2011-04-08T13:18:05+02:00 http://como.vub.ac.be/teaching:offlinetheses?rev=1302261485&do=diff Add the link to your thesis proposal here. This entire page will be moved to the “Teaching” on Friday, April 8, 2011 at 2:00pm. Thesis and Research training proposals * Learning Control for Quadrotor Helicopters * Channel Allocation in Wireless Sensor Networks * Optimization Algorithm Portfolios * Multi-agent Transfer Learning * 3D face recognition * Video based motion tracking and gesture recognition * Population based Reinforcement Learning * Multi-Agent Techniques for… text/html 2011-09-19T10:19:38+02:00 http://como.vub.ac.be/teaching:procdyn?rev=1316420378&do=diff Course code: WE-DINF-13865 This course will officially start in 2012. In 2010-2011, the course is divided into two courses * Dynamic Processes (MATHH406), and * Operational Research Applications (MATHH508). Teacher Professors: * Bertrand Mareschal for Operational Research Applications. * Pierre Louis Kunsch for Dynamic processes text/html 2009-10-06T11:45:10+02:00 http://como.vub.ac.be/teaching:project_mergesd?rev=1254822310&do=diff Context Subgroup Discovery (SD) is a knowledge discovery task belonging to the group of local pattern mining tasks. It was first introduced by Klosgen and Wrobel as a rule learning technique located at the intersection of predictive (or supervised analysis like classification) and descriptive induction (or unsupervised analysis like clustering). text/html 2009-10-06T16:21:57+02:00 http://como.vub.ac.be/teaching:project_music?rev=1254838917&do=diff Context Music information retrieval is the general denominator that englobes research about analyzing music data. Nowadays, lots of people work on audio signals and investigate style classification, querying in large databases, automatic tagging, pitch estimation, automatic transcription and so on. A complementary research area is working on computational analysis of symbolic music, looking into similar problems from a more musicological perspective. text/html 2009-10-05T14:38:24+02:00 http://como.vub.ac.be/teaching:project_rl?rev=1254746304&do=diff Reinforcement Learning (RL) describes a class of learning algorithms which can be used to let a learning agent optimize its behavior by trial and error. In the standard problem formulation, the agent is situated in an environment which can assume a number of states. Each time step the environment is in a certain state and the agent selects an action to perform. The result of this action is an immediate reward for the agent and a change in environment state. The agent's goal is to learn an action… text/html 2009-09-28T12:16:30+02:00 http://como.vub.ac.be/teaching:project_warehouse?rev=1254132990&do=diff General Context At the CoMo Lab, we are currently setting up a LEGO warehouse as a common test platform for our research. Our warehouse is an environment where Mindstorm robots are responsible for storing and retrieving a variety of objects from shelves. text/html 2008-05-30T22:24:53+02:00 http://como.vub.ac.be/teaching:qos?rev=1212179093&do=diff State of the art of the art The Internet Protocol (IP) has become the preferred protocol for the transmission of various applications (e.g. IPTV, video on demand, telemedicine, digital cinema, etc.). This has lead network operators to adopt Optical Networks as underlying transport technology. Optical networks provide the IP layer with a huge amount of capacity. As the traffic demand has become more dynamic and unpredictable, optical networks have been progressively equipped with intelligent age… text/html 2008-05-30T14:33:58+02:00 http://como.vub.ac.be/teaching:rl_for_transmission?rev=1212150838&do=diff Het onderwerp van dit eindwerk is gesitueerd in het domein van de lerende systemen. In vele industriële toepassingen zijn de exacte parameters van de gecontroleerde systemen slechts in beperkte mate gekend en bovendien ook veranderlijk doorheen de tijd, bv. door slijtage. Deze onzekerheid zorgt ervoor dat deze systemen meestal suboptimaal aangestuurd worden, wat leidt tot performantieverlies. Wanneer zulk systeem dezelfde cyclus echter meermaals doorloopt, kan de onzekerheid over het systeem opg… text/html 2008-05-30T14:34:27+02:00 http://como.vub.ac.be/teaching:robot?rev=1212150867&do=diff Introduction As part of CoMo’s Multi-agent learning research and teaching efforts we perform experiments using Lego Mindstorm robots. In these experiments one ore more robots are typically learning to perform a task in a shared arena. The purpose of this thesis is to help develop the arena used for performing these robot experiments. An important part of this goal is to be able to track robots’ movements. During the thesis you will be required to develop a video tracking system able to determi… text/html 2009-05-06T12:39:11+02:00 http://como.vub.ac.be/teaching:routing_and_wavelength_assignment_using_maximum_flow_based_algorithms?rev=1241606351&do=diff Context The advent of Dense WDM (DWDM) in optical networks has significantly increased the bandwidth available on a single fiber. It is expected that Wavelength Division Multiplexing (WDM) routing networks will be deployed to meet the high bandwidth needs of Internet traffic, especially in backbone networks for nationwide or global coverage. In a Multi-Domain framework, scalability is the main issue, since the amount of shared information has to be limited in order to prevent routing tables… text/html 2010-10-04T15:39:10+02:00 http://como.vub.ac.be/teaching:rts3ba?rev=1286199550&do=diff text/html 2010-10-04T15:06:39+02:00 http://como.vub.ac.be/teaching:slam3ba?rev=1286197599&do=diff text/html 2011-12-13T12:30:10+02:00 http://como.vub.ac.be/teaching:statml?rev=1323775810&do=diff IMPORTANT: The VUB course 'Adaptive Systems' will now be taught jointly with the ULB course 'Statistical Foundations of Machine Learning'. VUB Students who choose this course must register it using the 'Adaptive Systems' course code. Course code: WE-DINF-12892 text/html 2012-02-06T14:00:33+02:00 http://como.vub.ac.be/teaching:swarmint?rev=1328533233&do=diff Course code: WE-DINF-13863 / INFOH414 Lecturers M. Mauro Birattari and Professor Marco Dorigo Time and Place 2nd semester course. See detailed schedule here. (Note that ULB_week_number = VUB_week_number + 1) The course takes place every Thursday from 14h to 18h, except for March 22 and during the Eastern holidays. The course is in room S.UB4.329A (Solbosch campus) and in IRIDIA. See the schedule below for details. text/html 2012-03-19T19:53:15+02:00 http://como.vub.ac.be/teaching:tai?rev=1332183195&do=diff Teacher Professor Ann Nowe and Professor Hugues Bersini Teaching assistants: Bart Jansen and Peter Vrancx Time and Place The exam took place on Thursday, January 19, 2012 from 14h to 17h in room D005. Lectures When : From October 3rd, 2011. Mondays - 1pm till 3pm. text/html 2011-04-09T19:28:11+02:00 http://como.vub.ac.be/teaching:teaching_thesis_hrl?rev=1302370091&do=diff Background Strategy games offer a rich environment for testing AI approaches. The games include a number of challenges to which learning can applied such as: team work, adversarial play, exploration, division of labour and many more. This fact has led to a number of attempts to employ these games as a testing environment in research papers. The goal of this thesis is to investigate the potential application of Hierarchical reinforcement learning in an open source turn based strategy (TBS) game… text/html 2011-04-08T13:33:50+02:00 http://como.vub.ac.be/teaching:thesis_3d_faces?rev=1302262430&do=diff Context Face recognition is a very popular research area in the field of computer vision. Due to the nature of the problem it addresses questions to researchers from various disciplines such as image processing, machine learning, data mining or psychology. We are interested in face recognition methodologies that make use of physiological characteristics in order to accurately identify an individual. In general terms, the face recognition problem can be formulated as follows: given one or more… text/html 2011-04-26T10:55:58+02:00 http://como.vub.ac.be/teaching:thesis_adaptiveheuristics?rev=1303808158&do=diff Context A single decision-making agent acting rationally in a stationary environment has to maximize its total expected reward. This is an optimization problem which can be solved using reinforcement learning methods like Q-learning. In contrast, if decision-making agents interact strategically, i.e. the reward (or payoff) of one agent not only depends on its own decision but also of the decisions made by the others or in other words the agents are players in a game, the notion of rationality… text/html 2012-05-09T10:24:05+02:00 http://como.vub.ac.be/teaching:thesis_android_rl?rev=1336551845&do=diff Reinforcement learning (RL) is a machine learning technique which allows a learner to determine optimal behavior from trial and error interactions. By trying out possible actions and evaluating the results, the learner can derive a sequence of actions leading to its long term goal. RL has been successfully applied to a wide range of problems, including robotics control, scheduling, network routing, game AI, and more ... A key benefit of reinforcement learning, is that the methods can easily be m… text/html 2012-05-09T11:16:51+02:00 http://como.vub.ac.be/teaching:thesis_comp_bio?rev=1336555011&do=diff Three proposal are available in this domain: Computational Biology: Investigations into protein structure and function (Tom Lenaerts, Elisa Cilia) One of the main research tracks in the group is linked to questions related to the structure and function of proteins. Machine learning methods and advanced algorithms can assist in answering these questions.This is a short list of topics which we want to investigate: text/html 2012-05-15T13:10:48+02:00 http://como.vub.ac.be/teaching:thesis_diff_fames?rev=1337080248&do=diff Reinforcement Learning (RL) describes a class of learning algorithms which can be used to let a agent optimize its behavior by trial and error. In the standard problem formulation, the agent is situated in an environment which can assume a number of states. Each time step the environment is in a certain state and the agent selects an action to perform. The result of this action is an immediate reward for the agent and a change in environment state. The agent's goal is to map actions to states in… text/html 2011-04-09T19:27:51+02:00 http://como.vub.ac.be/teaching:thesis_dynscheduling?rev=1302370071&do=diff Background The goal of this thesis is to develop an Android application that allows dynamic scheduling of a number of activities, based on changing information that is received from a group of contacts. The idea is that multiple users can input a number of (possibly group) activities that need to be scheduled, together with a list of constraints. The application then creates an individual schedule for each user, which can be updated if new data about the activities becomes available (e.g. dela… text/html 2009-05-07T16:26:24+02:00 http://como.vub.ac.be/teaching:thesis_eiwitten?rev=1241706384&do=diff Context One of the major goals of molecular biology is to understand how protein chains fold into a unique 3-dimensional structure starting from their genetic code, and using an alphabet of only 20 amino acids. Given this knowledge, perhaps the most exciting prospect is the possibility of designing new proteins to perform designated tasks, an application that could prove to be of great importance in medicine and biotechnology. All of the 20 amino acids have in common a central carbon atom and … text/html 2009-05-08T13:27:21+02:00 http://como.vub.ac.be/teaching:thesis_gametheory?rev=1241782041&do=diff Context Game theory studies strategic interactions among players. The mathematical model of such an interaction is called a game. Literally, every entity in the world continuously engages in 'games'. Companies are competing for market share, countries convince others to counter environmental pollution, individuals invest in the stock market. Not just humans play these games, also animals, plants, cells, etc. text/html 2011-04-08T12:06:18+02:00 http://como.vub.ac.be/teaching:thesis_gesture_recognition?rev=1302257178&do=diff Context Within a broader framework of learning systems to behave in an “intelligent” manner, the recognition of specific patterns such as gestures, objects, activities etc. from sensors’ output is needed. In this proposal students are asked to develop a gesture recognition method capable to identify a set of predefined gestures from video signals. The correct recognition of gestures is needed so that the system could take the right action and to adapt in a dynamic environment. text/html 2012-05-09T11:21:49+02:00 http://como.vub.ac.be/teaching:thesis_heuristic?rev=1336555309&do=diff Context Reinforcement Learning allows an agent to learn optimal behaviour through trial-and-error interactions with its environment. By repeatedly evaluating possible actions in different situations, the agent can discover the consequences of his actions and select the optimal one. Normally, RL does not assume any prior information and the agent's environment is considered to be completely unknown. However, when dealing with specific learning problems expert human knowledge may be available. Th… text/html 2011-04-08T09:13:30+02:00 http://como.vub.ac.be/teaching:thesis_ma_transfer?rev=1302246810&do=diff Context Reinforcement learning is a mature technique for solving complex problems. Significant advances have been developed to speed up the learning process through the incorporation of domain knowledge, the decomposition of problems in subtasks, using a generalised state space representation or learning higher level actions, composed of multiple one step actions. The main drawback of all these techniques is however that each time the RL algorithm has to start the learning process from the beg… text/html 2011-04-26T11:13:32+02:00 http://como.vub.ac.be/teaching:thesis_metagenomics?rev=1303809212&do=diff Context Meta-genomics is a novel approach to DNA sequencing, in which the “genome” of complete microbial communities (e.g. the gut flora, ocean water samples) can be determined. Because the rapid advances in DNA sequencing technologies, this approach is yielding millions of genes which need to be analysed to understand how these ecosystem functions or how disturbances in our flora are linked to disease. One important step in this process is the functional annotation of genes and the identifica… text/html 2011-04-26T10:57:17+02:00 http://como.vub.ac.be/teaching:thesis_multiarmedbandit?rev=1303808237&do=diff Context [A one-armed bandit or slot machine.] A multi-armed bandit is a collection of slot machines or one-armed bandits. Each slot machine returns some random reward each time its lever or arm is pulled. Since slot machines provide different average payoffs, a gambler could increase his reward by choosing intelligently which arm to pull. text/html 2009-05-08T13:32:34+02:00 http://como.vub.ac.be/teaching:thesis_music?rev=1241782354&do=diff Context Music information retrieval is the general denominator that englobes research about analyzing music data. Nowadays, lots of people work on audio signals and investigate style classification, querying in large databases, automatic tagging, pitch estimation, automatic transcription and so on. However, there is still a smaller community that is interested in music analysis with a more musicological or cognitive point of view. In such a context, researchers deal with symbolic music digita… text/html 2012-05-09T16:06:43+02:00 http://como.vub.ac.be/teaching:thesis_perpetual?rev=1336572403&do=diff Context Combining an optimal product or system functionality with low resource consumption and minimal costs is a vision that fits well into recognized customer expectation trends. In reality often a trade-off has to be made between several criteria. For the temperature control of a heater, the economic cost of getting the room to a certain temperature might for example also be important for the user, next to the comfort of having the desired temperature. Therefore, a system will be developed, … text/html 2011-04-08T10:57:23+02:00 http://como.vub.ac.be/teaching:thesis_populationbasedrl?rev=1302253043&do=diff Context Current multi-agent learning approaches are limited to relatively small, semi-static systems, wherein the number of agents is fixed, the size is relatively small, and the environment changes at a slower rate than the time required to adapt the behaviour. As such, these limitations are at odds with what normally is expected of an adaptive system. The goal of this thesis is to develop a new model for learning in autonomous multi-agent systems that can handle large scale and highly dynami… text/html 2012-05-09T11:28:24+02:00 http://como.vub.ac.be/teaching:thesis_portfolios?rev=1336555704&do=diff Context Research in machine learning keeps producing more efficient algorithms for solving hard problems, however in most cases there is no single algorithm dominating in performance. This motivated an increasingly relevant corpus of research on metalearning [1], aimed at automating the combination of alternative algorithms into composite problem solvers which improve over the performance of each single component. Simple examples include the selection among alternative algorithms, and tuning o… text/html 2009-05-07T15:53:58+02:00 http://como.vub.ac.be/teaching:thesis_probo?rev=1241704438&do=diff Information * [Powerpoint presentation] * Probo website Contact Supervisors: Jelle Saldien and Kristof Goris text/html 2011-04-22T15:18:13+02:00 http://como.vub.ac.be/teaching:thesis_quadcopter?rev=1303478293&do=diff Background This thesis investigates the use of reinforcement learning for autonomous flight of 4-rotor helicopters (quadcopters). CoMo has recently acquired 2 Parrot AR Quadcopter Drones. These quadcopters can be controlled by users over WiFi, using iPhone or Android applications or custom developed software. Each helicopter has 2 camera’s in addition to a number of sensors (accelerometer, gyroscope, ultrasound) which can be read. During this thesis you will develop control software for the dr… text/html 2012-05-09T13:33:34+02:00 http://como.vub.ac.be/teaching:thesis_rcrl?rev=1336563214&do=diff Reservoir computing (RC) is a novel paradigm for training and designing recurrent neural networks. Recurrent neural networks extend the traditional feedforward network model by allowing cycles in the network’s structure. These cycles allow the network to use previous network outputs in its computations and effectively implement a type of memory. A downside of these enhanced network capabilities is that training the network weights becomes more complex. Reservoir computing removes this problem by… text/html 2011-04-26T11:05:18+02:00 http://como.vub.ac.be/teaching:thesis_robocup?rev=1303808718&do=diff Context Robocup.be is een nieuw innovatief testbed onder leiding van DSP Valley dat opgezet is rond het thema van embedded systemen en robotica. Dit initiatief kadert in een internationaal project dat AI en robotica onderzoek wil promoten door internationale voetbaltornooien voor robots te organiseren. Er zijn verschillende competities afhankelijk van de grootte van de robots: kleine, middelgrote of grote (op mensenmaat) robots (www.robocup.org). text/html 2011-04-26T10:57:47+02:00 http://como.vub.ac.be/teaching:thesis_scheduling?rev=1303808267&do=diff Context [Schedule for six jobs on six machines.] In general, scheduling is deciding which job or task is to be processed on which machine. Obvious examples of scheduling include creating timetables for nurses in a hospital or teaching activities in schools, allocating processor time to computer jobs, and allocating assembly lines or machines to produce different goods. All scheduling problems include some constraints, like: a machine can handle at most one job at a time. When all constraints … text/html 2009-05-07T16:01:26+02:00 http://como.vub.ac.be/teaching:thesis_sensornets1?rev=1241704886&do=diff Context The study of routing and MAC protocol design for energy efficient exploitation of sensor networks is a hot research topic nowadays. We like to make a contribution by introducing methods from the world of artificial intelligence in the design in order to increase the global performance of the nodes with respect to achieving their goal. This is called “collective intelligence”. This intelligence can be obtained by utilizing distributed learning algorithms and applying elements from game … text/html 2009-05-07T16:00:48+02:00 http://como.vub.ac.be/teaching:thesis_sensornets2?rev=1241704848&do=diff Context Contiki is an operating system for embedded systems that was developed by the leading research group SICS in Sweden. This OS works with a minimum of memory (a few kB of RAM and ROM) and is therefore highly suitable for systems with strong constraints on memory usage, like sensor nodes. If we group a number of such nodes to form a wireless networks, we call this a wireless sensor network. Besides memory usage, energy efficiency and reliability are “hot issues” in this interesting resear… text/html 2010-04-26T14:55:21+02:00 http://como.vub.ac.be/teaching:thesis_slam?rev=1272286521&do=diff Context Simultaneous localization and mapping (SLAM) is an important issue in mobile robotics. Mapping describes the problem of integrating sensor inputs (sonar, range finders,... ) into a coherent description (a map) of the environment. Localization is the problem of determining the robot's position withing the environment. These two problems often cannot be solved independently. Using observations to build a map requires information on location of the robot to determine where the observati… text/html 2010-04-29T14:18:10+02:00 http://como.vub.ac.be/teaching:thesis_soccer?rev=1272543490&do=diff Context Defining good soccer strategies for a team of agents is a straightforward problem if the other team is stationary. This is however never the case, so a team must be able to adapt its strategies if this is necessary. At CoMo a lot of research has been about adaptive state representations and observing other agents if necessary. text/html 2010-04-27T11:45:00+02:00 http://como.vub.ac.be/teaching:thesis_sparse?rev=1272361500&do=diff Context Reinforcement Learning (RL) describes a class of learning algorithms which can be used to let a learning robot optimize its behavior by trial and error. In the standard problem formulation, the robot is situated in an initially unknown environment which it can observe through its sensors. At regular intervals the robot makes a sensor observation (i.e. it determines the current environment state) and selects an action to perform. As a result of this action the robot may observe a change… text/html 2010-04-27T11:44:06+02:00 http://como.vub.ac.be/teaching:thesis_states?rev=1272361446&do=diff Context Reinforcement Learning (RL) allows an agent to learn optimal behavior through trial-and error interaction with its environment. By repeatedly evaluating possible actions in different circumstances (referred to as environment states) the agent can learn the optimal action in each case. However, when multiple agents are learning in a shared environment, these agents can interfere with each other's learning process. The most common solution to this issue is to always include information on… text/html 2012-05-09T11:08:14+02:00 http://como.vub.ac.be/teaching:thesis_trajectory?rev=1336554494&do=diff An important task in the control of autonomous vehicles, is to steer the vehicle along a predefined trajectory. Classical engineering control strategies can be used to accomplish this task, however, these methods typically rely on a static system model and can have problems adapting to changing conditions or model inaccuracies. The goal of this thesis is to investigate the application of machine learning methods to the problem of trajectory control. A key limitation in this setting is that the l… text/html 2010-10-04T16:21:03+02:00 http://como.vub.ac.be/teaching:tracking3ba?rev=1286202063&do=diff text/html 2010-05-14T15:02:16+02:00 http://como.vub.ac.be/teaching:using_machine_learning_algorithms_for_topology_aggregation_in_multi-domain_wdm_optical_networks?rev=1273842136&do=diff Context In fiber-optic communications, dense wavelength-division multiplexing (DWDM) is a technology which multiplexes multiple optical carrier signals on a single optical fiber by using different wavelengths (colours) of laser light to carry different signals. It is expected that DWDM routing networks will be deployed to meet the high bandwidth needs of Internet traffic, especially in backbone networks for nationwide or global coverage. In DWDM networks end-to-end users in the optical domain a… text/html 2010-05-14T15:17:20+02:00 http://como.vub.ac.be/teaching:using_machine_learning_algorithms_for_topology_aggregation_in_multi-domain_wdm_optical_networks.txt?rev=1273843040&do=diff Context In fiber-optic communications, dense wavelength-division multiplexing (DWDM) is a technology which multiplexes multiple optical carrier signals on a single optical fiber by using different wavelengths (colours) of laser light to carry different signals. It is expected that DWDM routing networks will be deployed to meet the high bandwidth needs of Internet traffic, especially in backbone networks for nationwide or global coverage. In DWDM networks end-to-end users in the optical domain a… text/html 2008-05-30T14:34:12+02:00 http://como.vub.ac.be/teaching:wii?rev=1212150852&do=diff Introduction One of the main drives behind the success of the Nintendo Wii entertainment console is the innovative WiiMote controller. Each WiiMote is equipped with bluetooth, accelerometers and infrared tracking. These extensive capabilities have led to a growing number of hacking projects based on the WiiMote. One of these projects is the WiiWhiteboard, developed by Johnny Chung Lee. In this application a WiiMote is used to track the movements of an infrared LED. This LED can then be used a… text/html 2012-05-09T16:22:42+02:00 http://como.vub.ac.be/teaching:wsn_ml?rev=1336573362&do=diff Context Wireless sensor networks (WSN) form an emerging class of networks able to monitor environments with high spatiotemporal accuracy. The network is composed of tiny devices known as wireless sensors or motes, endowed with a microprocessor, a memory, a radio, a battery, and one or more sensors such as temperature, humidity, light or sound sensors. The figure below gives an illustration of a typical wireless sensor platform used in research (TMote), and of an integrated silicon design (Depu…