Change search
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf
Synthetic Meta-Learning:: Learning to learn real-world tasks with synthetic data
KTH, School of Electrical Engineering and Computer Science (EECS).
2019 (English)Independent thesis Advanced level (degree of Master (Two Years)), 20 credits / 30 HE creditsStudent thesisAlternative title
Syntetisk metainlärning: : Lära sig att lära verkliga uppgifter med syntetisk data (Swedish)
Abstract [en]

Meta-learning is an approach to machine learning that teaches models how to learn new tasks with only a handful of examples. However, meta-learning requires a large labeled dataset during its initial meta-learning phase, which restricts what domains meta-learning can be used in. This thesis investigates if this labeled dataset can be replaced with a synthetic dataset without a loss in performance. The approach has been tested on the task of military vehicle classification. The results show that for few-shot classification tasks, models trained with synthetic data can come close to the performance of models trained with real-world data. The results also show that adjustments to the data-generation process, such as light randomization, can have a significant effect on performance, suggesting that fine-tuning to the generation process could further improve performance.

Abstract [sv]

Metainlärning är en metodik inom maskininlärning som gör det möjligt att lära en modell nya uppgifter med endast en handfull mängd träningsexempel. Metainlärning kräver dock en stor mängd träningsdata under själva metaträningsfasen, vilket begränsar de domäner där metodiken kan användas. Detta examensarbete utreder huruvida syntetisk bilddata, som genererats med hjälp av en simulator, kan ersätta verklig bilddata under metainlärningsfasen. Metoden har utvärderats på militär fordonsklassificering. Resultaten visar att för bildklassificering med 1–10 träningsexempel per klass kan en modell metainlärd med syntetisk data närma sig prestandan hos en modell metainlärd med riktig data. Resultaten visar även att små ändringar i genereringsprocessen, exempelvis graden av slumpmässigt ljus, har en stor inverkan på den slutgiltiga prestandan, vilket ger hopp om att ytterligare finjustering av genereringsprocessen kan resultera i ännu fler prestandaförbättringar.

Place, publisher, year, edition, pages
2019. , p. 64
Series
TRITA-EECS-EX ; 2019:640
National Category
Computer and Information Sciences
Identifiers
URN: urn:nbn:se:kth:diva-264919OAI: oai:DiVA.org:kth-264919DiVA, id: diva2:1375764
External cooperation
Swedish defence research agency
Supervisors
Examiners
Available from: 2019-12-09 Created: 2019-12-06 Last updated: 2019-12-09Bibliographically approved

Open Access in DiVA

fulltext(787 kB)32 downloads
File information
File name FULLTEXT01.pdfFile size 787 kBChecksum SHA-512
def25d5107f1f0b3266ed009e1b1ca05f64c361eb04615fc5f118791112fdea1555769a49575136d51773c22755ef97716a5eb887410de6bc77d2ed554005523
Type fulltextMimetype application/pdf

By organisation
School of Electrical Engineering and Computer Science (EECS)
Computer and Information Sciences

Search outside of DiVA

GoogleGoogle Scholar
Total: 32 downloads
The number of downloads is the sum of all downloads of full texts. It may include eg previous versions that are now no longer available

urn-nbn

Altmetric score

urn-nbn
Total: 74 hits
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf