0x010088F6312065A70A44B630A6848498EAD80093CFEAFE161DC94D83C7C5C9A5A748B8

<ul><li>Suggestion: Waiting times in the exercise sessions are pretty long (but the effort was appreciated)</li><li>Action: We arranged at least three TAs for each exercise session (compared to 2 TAs last year).</li></ul> <ul><li>Suggestion: Having the deadlines around noon is easier for the students</li><li>Action: We set the deadlines at 22:00 in this year instead of 23:59 in last year.</li></ul> <ul><li>Suggestion: Maybe the individual project could be changed to something more related to new themes, which is special for ML? We have already coded K-means/decision trees in other courses. Maybe the assignment could be more special just for ML.</li><li>Action: We updated the task and datasets. More tasks could be added next year.</li></ul>

The student feedback for the course is generally positive. Students appreciate the clear distribution of information through Blackboard and Piazza. They also value the openness of the coordinators to feedback. The exercise sessions are found to be very helpful, and the availability of numerous study-group sessions is well-received. Additionally, students acknowledge that the course provides a lot of practical experience.

Improvement plan according to the suggestions from students <ul><li>Suggestion: The individual assignment should have clearer requirement.</li><li>Plan: We will improve the requirement documentation.</li></ul> <ul><li>Suggestion: Split the machine learning course.</li><li>Plan: The course will be split into three parts since 2024. A beginner course for machine learning basics will be introduced. TDT4173 will then focus on more advanced topics on realistic machine learning applications. In addition, the theory module course TDT05 will teach how to develop new machine learning tools.</li></ul> <ul><li>Suggestion: Distribute exercise sessions more evenly throughout the week.</li><li>Plan: We will notify the TimePlan about this.</li></ul> <ul><li>Suggestion: Improve TA preparedness for projects.</li><li>Plan: All TAs will be more involved in virtual team development to better understand the project tasks and data.</li></ul> <ul><li>Suggestion: Integrate theoretical understanding with practical application.</li><li>Plan: We will incorporate additional lectures to demystify the theory behind modern machine learning tools.</li></ul> <ul><li>Suggestion: Provide practical sessions on model implementation and tool selection.</li><li>Plan: We will expand the content to include more practical guidance on models, implementation, preprocessing, and tool selection.</li></ul> <ul><li>Suggestion: Reorganize syllabus to align with the final assessment.</li><li>Plan: We will reorder the topics to introduce more relevant material earlier, aiding students in their final assessments.</li></ul> <ul><li>Suggestion: Offer examples of good solutions from previous years.</li><li>Plan: We will showcase exemplary solutions from previous years and discuss what made them successful, along with common pitfalls to avoid.</li></ul>

NTNU:TIA:REST

Institutt for datateknologi og informatikk

Department of Computer Science

Engelsk

Mer om vurdering

Emneevalueringen består av to deler. (1) Hver student må først bestå en individuell oppgave (IA) ca. en måned etter kursstart. Hver student kan få et nytt forsøk til å bestå IA, men studenten får et fradrag (-5%) i emnepoengene. Studenter som stryker i IA på begge forsøkene vil få karakteren F eller Ikke bestått. (2) Kun de som består IA kan fortsette til kursprosjektet. Prosjektet vurderes for hele teamet (hvert team består av inntil tre studenter). Prosjektpoengene tilsvarer basispoeng (maks. 100 % og min. 41 %) pluss potensielle prosjektfradrag (fra 0 % til -17 %). Basispoengene er proporsjonale med antallet virtuelle lag (VT) som er beseiret av studentlaget med hensyn til predikasjonsprestasjon. Lærerne og undervisningsassistentene forbereder VT-ene. Hvis et studentlag ikke kan beseire noen VT, vil teammedlemmene mislykkes i prosjektet og dermed kurset. De potensielle fradragene inkluderer en sen innlevering (innen tre dager) og manglende dokumentering av nøkkelkomponenter i maskinlæringspraksis.Kurspoengene vil til slutt avrundes til bokstavkarakter etter NTNUs standardintervaller. Dersom studenten får F eller Ikke bestått som sluttkarakter, må studenten ta hele emnet på nytt.

Anbefalte forkunnskaper

Emnene TDT4136 Introduksjon til kunstig intelligens, TDT4171 Metoder i kunstig intelligens eller tilsvarende.

Faglig innhold

Emnet introduserer prinsipper og metoder for automatisert læring i datasystemer. Undervisningen fokuserer på konsepter, teknikker og analysemetoder knyttet til datadrevet læring og generalisering til nye datapunkter. Tidsseriemetoder, ensemblemetoder, dyp læring, statistiske generaliseringer og reproduserbarhet er også inkludert. Det legges vekt på de forskjellige læringsmetodenes styrker og svakheter. Læringsmetoder fra case-based reasoning blir også beskrevet. I tillegg til teoretisk kunnskap gir emnet praktisk veiledning for utvikling av et maskinlæringssystem.

Kursmateriell

Lærebøker:<ul><li>Tom Mitchell: Machine learning, McGraw Hill, 1997.</li><li>Christopher M. Bishop: Pattern Recognition and Machine Learning, 2006</li></ul>Utvalgte artikler.

Læringsutbytte

Emnet har som mål å introdusere overordnede prinsipper innenfor maskinlæringsmetoder og gi forståelse av nøkkelelementer for praktisk anvending av maskinlæring.

Læringsformer og aktiviteter

Forelesninger, gruppearbeid, kollokvier og selvstudium.

Maskinlæring

Industriell økonomi

Informasjonssikkerhet

Informatikk

Psykologi

Statistikk

Teknologiske fag

Datateknologi - masterstudium (5-årig)

Kybernetikk og robotikk - masterstudium (2-årig)

Informasjonsteknologi - doktorgrad (ph.d.)

Telematikk - doktorgrad (ph.d.)

Industriell kybernetikk - masterstudium (2-årig)

Digital Infrastructure and Cyber Security (Master's Programme)

Teknisk kybernetikk - doktorgrad (ph.d.)

Kybernetikk og robotikk - masterstudium (5-årig)

Datateknologi - masterstudium (2-årig)

Master of Science in Informatics

Web/Lists(guid'8b698a2d-6ad0-4f7d-b3be-219cc1979f02')/Items(26493)/FieldValuesAsText

Below we review some suggestions by the students and present our action points for improvement in the future.

Suggestion: Better with own preference in teaming
Action: Will do in next year. Students will be able to form their team by themselves.

Suggestion: Maybe do the lab activities in computer lab rooms so that running the codes is faster.
Action: Will try to book a suitable computer classroom for this purpose.

Suggestion: The assessment form is very focused on one specific task. One task seems to be easier than the other. Perhaps two projects are better.
Action: Will consider requiring both tasks next year

Suggestion: Perhaps the assessment could be a bit broader. Some efforts should be taken into account in addition to the prediction score.
Action: Will include more grading items; not only with deduction points, but also with some bonus items.

Suggestion: The description of the course grading does not match the actual grading criteria
Action: The description on the NTNU page is obsolete. Will update the page next year.

Suggestion: Maybe have some quick theoretical assignments in order to study better also the theory
Action: Will add one or more individual assignments on theory

Suggestion: Better to move the deadline one or two weeks earlier to avoid overlap with the exam week
Action: Will do that

Suggestion: Make the requirements more apparent and consistent
Action: The competitive assessment method was run for the first time. We will try hard to improve the implementation and make the information consistent.


Tiltak

Development plan for TDT4173 in 2022

Below are the actions corresponding to the suggestions from last year. <ul><li>Suggestion: Better with own preference in teaming</li><li>Action: Done. Students formed project groups by themselves.</li></ul> <ul><li>Suggestion: Maybe do the lab activities in computer lab rooms so that running the codes is faster.</li><li>Action: We have provided virtual machines for students as extra computing resources. We have not received any negative feedback about computing cost.</li></ul> <ul><li>Suggestion: The assessment form is very focused on one specific task. One task seems to be easier than the other. Perhaps two projects are better.</li><li>Action: All students participated in the same project.</li></ul> <ul><li>Suggestion: Perhaps the assessment could be a bit broader. Some efforts should be taken into account in addition to the prediction score.</li><li>Action: We have made the grading guide more detailed. Other efforts have been taken into account. We are considering some bonus items in the coming years.</li></ul> <ul><li>Suggestion: The description of the course grading does not match the actual grading criteria</li><li>Action: The description on the NTNU page has been updated.</li></ul> <ul><li>Suggestion: Maybe have some quick theoretical assignments in order to study better also, the theory</li><li>Action: Some theoretical parts have been added in the lectures. In the future, the course will focus more on machine learning basics and practices. More comprehensive theories will be moved to another course on advanced machine learning.</li></ul> <ul><li>Suggestion: Better to move the deadline one or two weeks earlier to avoid overlap with the exam week</li><li>Action: We have moved the deadline one week earlier without overlapping with the exam week.</li></ul> <ul><li>Suggestion: Make the requirements more apparent and consistent</li><li>Action: The requirements have been announced at the beginning. We have kept them consistent.</li></ul>

Most students welcomed the new competitive assessment method. Many students have said that they have learned a lot through the individual assignment and the group project work. The application in practice helps them understand the theoretical concepts taught during the lectures. See the attached students' feedback.

Below we review some suggestions by the students and present our action points for improvement in the future. <ul><li>Suggestion: Waiting times in the exercise sessions are pretty long (but the effort was appreciated)</li><li>Action: Will add more TAs for the course</li></ul> <ul><li>Suggestion: Having the deadlines around noon is easier for the students</li><li>Action: Will do that in the next years</li></ul> <ul><li>Suggestion: Maybe the individual project could be changed to something more related to new themes, which is special for ML? We have already coded K-means/decision trees in other courses. Maybe the assignment could be more special just for ML.</li><li>Action: Will create new tasks for the individual assignment next year</li></ul>

Emnet behandler prinsipper og metoder for hvordan datasystemer selv kan oppdatere sin kunnskap og problemløsningsevne. Klassiske metoder for maskinell læring basert på observerte data, samt læring som også utnytter eksisterende kunnskap gjennomgås og analyseres. Hovedvekten legges på symbolprosesserende metoder, der eksplisitte begreper og sammenhenger læres. Andre metoder som omhandles er bl.a. statistiske generaliseringer, tidsserier metoder, ensemble metoder, og dyp læring. Det legges vekt på de enkelte læringsmetoders sterke og svake sider, relatert til ulike mål for læringen. I case-basert resonnering integreres læring og problemløsning ved at konkrete problemer som løses tas vare på og gjenbrukes ved løsning av nye problemer, jfr. CBR-syklusen. Numeriske og kognitive modeller for similaritetsvurdering gjennomgås, og ulike systemarkitekturer diskuteres. Videre behandles metoder der både case-spesifikk og generell kunnskap integreres i lærings- og problemløsningsfasene.

Lærebok:<ul><li>Tom Mitchell: Machine learning, McGraw Hill, 1997.</li><li>Michael M. Richer and Rosina Weber: Case-Based Reasoning, Springer, 2013.</li></ul>Utvalgte artikler.

Emnet skal gi en innføring i prinsipper og metoder for hvordan datasystemer kan lære av egen erfaring. Det legges vekt på en fundamental forståelse av maskinlæring generelt, som en bakgrunn for ulike spesifikke metoder. I case-basert resonnering vektlegges forståelsen av hvordan maskinlæring og problemløsning integreres.

Web/Lists(guid'8b698a2d-6ad0-4f7d-b3be-219cc1979f02')/Items(20997)/FieldValuesAsText

The course basically follows the same organization as last year, with one responsible teacher, three other lecturers from the department, two guest lecturers, and six teaching assistants.There is a major change in the course project and its grading scheme. Last year, students could propose their own work for the course project. It was hard to grade their work due to the large diversity of their proposals.Machine learning is a science about prediction. Therefore we designed a new competitive assessment method. The students first completed an individual assignment. After passing the assignment, the students proceeded to a group project (3 students in a group). Each group could choose one of two given real-world machine learning tasks in the project. The students used provided training data to fit their machine learning models and then used them to predict for the testing data. We graded their performance by comparing their predictions with the held-out ground truths. Teams with more accurate predictions would get more project points. In the implementation, the teaching team prepared a collection of virtual teams in the competition, where the students' project points are proportional to the number of virtual teams they defeated. Besides the competition ranking, the teaching team graded the work by examining the necessary project steps in the report.

Most students welcomed the new competitive assessment method. Many students have said that they have learned a lot through the individual assignment and the group project work. The application in practice helps them understand the theoretical concepts taught during the lectures.

During the mid-term reference group meeting, we received the following suggestions and acted accordingly: Suggestion: More lab hours are needed.Action: Added Thursday 10-12 Suggestion: Use separate pages in Blackboard for learning materialsAction: Separated folders by topics/lectures Suggestion: Project requirements and guidelines in one PDF or pageAction: Added a PDF on submission and grading guidelines Suggestion: Be more consistent in course timing and breakAction: Have notified all lecturers

Mappevurdering er grunnlaget for karakteren i emnet. Mappen inkluderer innlevering av en metodeoppgave (35%), individuell oppgave (vurdering for en annen studentoppgave/quiz/osv) (15%) og et prosjekt (50%). Resultatene for delene er gitt i %, mens hele mappen tildeles bokstavkarakter. I tilfelle studenten får F / Ikke bestått som sluttkarakter, må studenten ta emnet på nytt i sin helhet.

Forelesninger, gruppearbeid, kollokvier, selvstudium og øvinger.

Communication Technology (Master's Programme)

Web/Lists(guid'8b698a2d-6ad0-4f7d-b3be-219cc1979f02')/Items(15457)/FieldValuesAsText

Maskinlæring - TDT4173