Grades 2026-01-22
Result discussion: 2026-01-23T13:00 (local time). Same Webex meeting as the lab.
The make-up exam will take place as scheduled, but in C112, not C308, as is listed.

General Information


Prezente minime la laborator: 70%. Calcul nota finala: (Lab + Exam) / 2 >= 5 .
In cazul in care una dintre notele Lab sau Exam este sub 5, se va face o trunchiere (si nu o rotunjire) a mediei (Laborator + Examen) / 2.
Toate temele trebuie trimise pe e-mail, in urmatoarele conditii:
  • Subiect: [NIM]_Nume_Prenume_MOC1_H1p
  • Atasament: format .zip
  • Atasat: codul sursa (inclusiv fisiere de descriere a proiectului, gen Makefile), fara binare compilate (!).
  • Atasat: raport in format .pdf, realizat in LaTeX. Folositi un template de jurnal / conferinta. E.g.: IEEE conference template . Folositi limba Engleza. Nu atasati sursele raportului.
  • O bibliografie detaliata este obligatorie.

Minimum number of attendances for the lab: 70%. Final grade formula: (Lab + Exam) / 2 >= 5 .
If either the Exam grade or Lab grade is below 5, the average (Lab + Exam) / 2 will be rounded down (truncation), instead of being rounded to the nearest integer.
All homework needs to be sent by e-mail, with the following requirements:
  • Subject: [NIM]_Surname_GivenName_MOC1_H1p
  • Attachment: only .zip archives
  • Attach: the source code of your project, including project files (e.g. project files for a certain IDE, or a Makefile), without any compiled binaries (!).
  • Attach: the report, in .pdf format, typeset in LaTeX. Use a journal / conference template. E.g.: IEEE conference template . Do not attach any source files for the report.
  • A detailed bibliography is mandatory.


Fisa disciplinei

Homework


Nota la laborator se obtine in urma rezolvarii urmatoarelor teme, cu formula:
[max( (0.8 * H1 + 0.2 * H1'), H1Conf) + max( (0.8 * H2 + 0.2 * H2'), H2Conf) + H3] / 3


H1Conf
Studentii care au studiat in anul II disciplina Algoritmi Genetici, in locul Temei H1 vor lucra pe o problema la alegere de la competiile organizate in cadrul conferintelor IEEE CEC si GECCO (sau altele, pe care le puteti propune.) Termen limita: saptamana 6.

Tema H1
  • Implementati un Algoritm Genetic cu reprezentare binara pentru optimizare numerica. Hibridizati Algoritmul Genetic cu euristica Hill Climbing.
  • Functiile de test vor fi: Rastrigin's, Griewangk's, Rosenbrock's, Michalewicz's
  • 2, 30 si 100 de dimensiuni.
  • Precizie: minim 5 zecimale dupa virgula.
    Termene limita:
  • algoritmul HC: saptamana 3
  • algoritmul genetic, hibridizarea si alte elemente de optimizare: saptamana 5.

Tema H1' Informatii utile aici
Se da functia f=x3-60x2+900x+100. Se doreste determinarea punctului de maxim atunci cand x ia valori in intervalul de intregi[0, 31].
Pe intervalul continuu [0, 31] functia este uni-modala, avand un singur punct de maxim care se obtine in valoarea 10.
Se doreste maximizarea functiei utilizand un algoritm Hill Climbing in care o solutie candidat este reprezentata pe 5 biti (32 valori posibil a fi reprezentate, ceea ce genereaza o cautare in spatiul numerelor intregi cuprinse intre 0 si 31). Vecinatatea consta in toate sirurile aflate la distanta Hamming 1. Studiati cum devine relieful fitness atunci cand se utilizeaza varianta first improvement si best improvement ale algoritmului Hill Climbing. In acest sens specificati punctele de maxim local si bazinele lor de atractie (bazin de atractie = multimea de puncte initiale pentru care cautarea ne duce spre acelasi optim).
(rezolvarea se poate face pe hartie, +fotografiat, verificat claritatea fotografiei, trimis pe e-mail).
Termen limita: saptamana 6. Daca aveti H1 si H1', le puteti prezenta impreuna in saptamana a 6-a.
Tema H2
Implementati Particle Swarm Optimisation pentru rezolva tema H1. Analizati influenta parametrilor asupra convergentei metodei.
Termen limita: saptamana 9.

Tema H2'
Hibridizati PSO pentru a ii imbunatati performanta. Realizati o analiza comparativa a metodelor implementate (H1-H2-H2').
Termen limita: saptamana 9.

Tema H2Conf
Continuat lucrul de la H1Conf (sau incepeti sa lucrati la problema de competitie, daca ati facut H1). Trebuie sa aveti rezultate comparabile cu alti participanti (Top 5) din competitia aleasa.
Termen limita: saptamana 9.
Tema H3
    Implementati doua metode la alegere (GA, ACO, PSO,CA-Automate Celulare) pentru a rezolva una dintre problemele de mai jos:
  • Problema MinMax Multiple TSP
  • Problema colorarii grafurilor (instante test)
  • o alta problema propusa de student

    Termene limita:
  • O metoda: saptamana 11
  • Ambele metode: saptamana 12

Tema H3Conf
Continuati lucrul la H2Conf. Fie va inscrieti la o competitie, fie obtineti rezultate mai bune decat locul 2 de la competitie, din clasamentul cel mai recent.
Termen limita: saptamana 12.


The Lab points are obtained by doing the following homework, using this formula:
[max( (0.8 * H1 + 0.2 * H1'), H1Conf) + max( (0.8 * H2 + 0.2 * H2'), H2Conf) + H3] / 3


H1Conf
Students who have studied Genetic Algorithms will choose one problem from a conference competition, like IEEE CEC or GECCO (or others, proposed by the student). Deadline: week 6.

H1
  • Implement a Genetic Algorithm with binary numerical representation. Hybridise the algorithm with the Hill Climbing heuristic.
  • Use the following test functions: Rastrigin's, Griewangk's, Rosenbrock's, Michalewicz's
  • 2, 30 and 100 dimensions.
  • Precision: at lest 5 precise decimal places.
    Deadlines:
  • Just a Hill Climber: week 3.
  • The Genetic Algorithm, the HC hybridisation, and other optimisation elements: week 5.

H1'
Given the function: f=x3-60x2+900x+100 , we want to study how its maximum is found on the integer interval [0, 31].
On this interval, the function is uni-modal, having a single maximum point in 10.
Study how a Hill Climbing algorithm maximises the value of the function. The HC will have a candidate solution of 5 bits (just enough for the 32 integers between 0 and 31). The neighbourhood is all strings at Hamming distance 1. Study how the fitness landscape changes between the first improvement and best improvement HC variants.
Specify the local maxima points and their attraction basins (attraction basin: the set of initial points for which the search finishes in the same local optimum).
(you may send a .pdf, or you may write on paper, photograph it, check the clarity and readability, and e-mail that).
Deadline: week 6. If you have H1', you may present H1 + H1' together, in week 6.
H2
Implement a Particle Swarm Optimisation algorithm to solve H1. Analyse how parameters influence algorithm convergence.
Deadline: week 9.

H2'
Hybridise PSO to improve its performance. Compare the methods you've implemented (H1-H2-H2').
Deadline: week 9.

H2Conf
Continue working on the competition chosen at H1Conf (or start working on the competition problem, if you worked on H1 before). You need to have results comparable to other participants (Top 5).
Deadline: week 9.
H3
    Implement two methods (Out of GA, ACO, PSO, CA - Cellular Automata) to solve one of the following problems:
  • The MinMax Multiple TSP problem.
  • The Graph Colouring problem(test instances. If the link is down, choose another database.)
  • Other problems proposed by students.
    Deadlines:
  • One method: week 11.
  • Both methods: week 12.

H3Conf
Continue working on H2Conf. Either apply to the competition, or get better results than the second place in the most recent ranking.
Deadline: week 12.

Lectures Slides

Lectures slides link (updated 2025-01-09):
nimSlides2025-2026.7z . The archive password is the same as the Webex sessions password.

Python Visual OpenGl Algorithms(BF, RSA, BIHC, GA, GA+BIHC, PSO)

Download:
solverGl.py

Data Mining lecture slides:
dataMining2024-2025.7z