AI Paper Atlas

Research map / Learning theory and algorithms / Matrix and tensor methods

Multitask Spectral Learning of Weighted Automata

Guillaume Rabusseau, Borja Balle, Joelle Pineau

NeurIPS 2017 · Learning theory and algorithms › Matrix and tensor methods · approach: High-dimensional estimation theory · 7 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Add multitask learning (MT-SL) with all other languages as related tasksSL · UNIDEP train size 100perplexity rel. improvement: 7.0744Improved
Restrict related tasks to the 4 closest languagesMT-SL · UNIDEP train size 1000perplexity rel. improvement: 3.2879 -> 4.4049Improved
Restrict related tasks to the 4 closest languagesMT-SL · UNIDEP train size 5000perplexity rel. improvement: 3.4187 -> 2.9689No clear effect
Restrict related tasks to the 4 closest languagesMT-SL · UNIDEP train size 100WER rel. improvement: 1.4919 -> 2.0883Improved
Add projection step (ablation)MT-SL-noproj · UNIDEP all data, all languagesWER rel. improvement: −9.600 -> 1.4932Improved
Add projection step (ablation)MT-SL-noproj · UNIDEP all data, all languagesperplexity rel. improvement: 0.6958 -> 3.1574Improved
Bag all languages' data into one SL training set instead of MT-SLSL · UNIDEP train size 100, all languagesperplexity rel. improvement: −19.00Hurt
Bag data of 4 closest languagesSL · UNIDEP all data, 4 closestWER rel. improvement: −3.588Hurt
Use MT-SL with 2, 4 or 8 related tasks instead of 2MT-SL · synthetic dS=10, dT=0 and 5perplexity and WERImproved
Use MT-SL when tasks are loosely related (dS = dT = 10)SL · syntheticperplexity and WERNo clear effect
Use MT-SL on small task-specific dimension (dT = 0, 5)SL · syntheticperplexity and WERImproved
Use MT-SL with 4 closest languagesSL · Basque, train size 500WER: ∼76% -> ∼70% -> ∼65%Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy