AI Paper Atlas

Research map / Label-efficient and robust learning / Recognition and long-tail

Human-Guided Complexity-Controlled Abstractions

Andi Peng, Mycal Tucker, Eoin M. Kenny, Noga Zaslavsky, Pulkit Agrawal, Julie Shah

NeurIPS 2023 · Label-efficient and robust learning › Recognition and long-tail · approach: Contrastive representation theory · 1 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
Use VQ-VIB_C instead of beta-VAE or VQ-VIB_N encodersfinetuning encoder · FashionMNIST 3-way, k=1accuracy: 40% and 55% -> 70%Improved
Reduce representation complexity (higher MSE) up to a pointVQ-VIB_C · FashionMNIST 3-way, k=1accuracy: MSE < 0.055 positive correlation, worse beyondMixed
Use VQ-VIB_C instead of beta-VAEfinetuning encoder · CIFAR100 2-way, k=1accuracy: 50% -> 70%Improved
Increase finetuning examples per class from 1 to 50encoders · CIFAR100 2-wayaccuracy gapNo clear effect
Compress representations on 20-way taskVQ-VIB_C · CIFAR100 20-way, k=1accuracyHurt
Increase number of combined quantized vectors n (1 to 2, 4) (ablation)VQ-VIB_C · FashionMNIST, iNatfinetuning accuracy: n=1 bestHurt
Penalize entropy instead of complexity (ablation)VQ-VIB_C and VQ-VIB_N · FashionMNISTfinetuning accuracyImproved
Use VQ-VIB_C instead of VQ-VIB_N under entropy regularization (ablation)encoder · FashionMNISTfinetuning accuracyImproved
Choose most complex encoder when finetuning data is large (k=50)encoder selection · FashionMNIST, CIFAR100, iNataccuracyNo clear effect
Select encoder by validation-set accuracy with small k and vencoder selection · FashionMNIST k=2, v=1accuracy: 73% vs over 80%Hurt
Increase validation set and kencoder selection · FashionMNISTaccuracy: 0.73 -> 0.97Improved
Increase finetuning data to 1010-way iNat with large kVQ-VIB_C vs beta-VAE · iNat 1010-wayaccuracyHurt

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