GDumb: A simple approach that questions our progress in continual learning

GDumb: A simple approach that questions our progress in continual learning

We discuss a general formulation for the Continual Learning (CL) problem for classification—a learning task where a stream provides samples to a learner and the goal of the learner, depending on the samples it receives, is to continually upgrade its knowledge about the old classes and learn new ones...

Deskribapen osoa

Xehetasun bibliografikoak
Egile Nagusiak:	Prabhu, A, Torr, PHS, Dokania, PK
Formatua:	Conference item
Hizkuntza:	English
Argitaratua:	Springer International Publishing 2020

Antzeko izenburuak

RanDumb: a simple approach that questions the efficacy of continual representation learning
nork: Prabhu, A, et al.
Argitaratua: (2025)

RanDumb: random representations outperform online continually learned representations
nork: Prabhu, A, et al.
Argitaratua: (2025)

Continual learning in low-rank orthogonal subspaces
nork: Chaudhry, A, et al.
Argitaratua: (2020)

An embarrassingly simple approach to zero-shot learning
nork: Romera-Paredes, B, et al.
Argitaratua: (2015)

Progressive skeletonization: trimming more fat from a network at initialization
nork: de Jorge, P, et al.
Argitaratua: (2020)

Computationally budgeted continual learning: what does matter?
nork: Prabhu, A, et al.
Argitaratua: (2023)

Discovering class-specific pixels for weakly-supervised semantic segmentation
nork: Chaudhry, A, et al.
Argitaratua: (2017)

Rapid adaptation in online continual learning: are we evaluating it right?
nork: Hammoud, HAAK, et al.
Argitaratua: (2024)

A continuous mapping for augmentation design
nork: Tian, K, et al.
Argitaratua: (2021)

Real-time evaluation in online continual learning: a new hope
nork: Ghunaim, Y, et al.
Argitaratua: (2023)

FedSR: a simple and effective domain generalization method for federated learning
nork: Nguyen, AT, et al.
Argitaratua: (2023)

Continual learning with tiny episodic memories
nork: Chaudhry, A, et al.
Argitaratua: (2019)

Using hindsight to anchor past knowledge in continual learning
nork: Chaudhry, A, et al.
Argitaratua: (2021)

Lifelong benchmarks: efficient model evaluation in an era of rapid progress
nork: Prabhu, A, et al.
Argitaratua: (2024)

Mirror Descent view for Neural Network quantization
nork: Ajanthan, T, et al.
Argitaratua: (2021)

Sample-dependent adaptive temperature scaling for improved calibration
nork: Joy, T, et al.
Argitaratua: (2023)

Using mixup as a regularizer can surprisingly improve accuracy and out-of-distribution robustness
nork: Pinto, F, et al.
Argitaratua: (2023)

Mix-MaxEnt: improving accuracy and uncertainty estimates of deterministic neural networks
nork: Pinto, F, et al.
Argitaratua: (2021)

On using focal loss for neural network calibration
nork: Mukhoti, J, et al.
Argitaratua: (2020)

Calibrating deep neural networks using focal loss
nork: Mukhoti, J, et al.
Argitaratua: (2020)

Placing objects in context via inpainting for out-of-distribution segmentation
nork: De Jorge, P, et al.
Argitaratua: (2024)

Scalable continual deep learning with computational cost considerations
nork: Prabhu, AP
Argitaratua: (2025)

Continual learning on a diet: learning from sparsely labeled streams under constrained computation
nork: Zhang, W, et al.
Argitaratua: (2024)

Are vision transformers always more robust than convolutional neural networks?
nork: Pinto, F, et al.
Argitaratua: (2021)

An impartial take to the CNN vs transformer robustness contest
nork: Pinto, F, et al.
Argitaratua: (2022)

TV, science and kids : teaching our children to question /
nork: Hays, Kim
Argitaratua: (1984)

STEER: simple temporal regularization for neural ODEs
nork: Ghosh, A, et al.
Argitaratua: (2020)

The service conception: Just one simple question
nork: Kirby, N
Argitaratua: (2017)

Progress in our understanding of the pathogenesis of ankylosing spondylitis
nork: Simone, D, et al.
Argitaratua: (2018)

Make some noise: reliable and efficient single-step adversarial training
nork: de Jorge, P, et al.
Argitaratua: (2023)

Raising the bar on the evaluation of out-of-distribution detection
nork: Mukhoti, J, et al.
Argitaratua: (2023)

Productivity in wood machining processes - a question of simple economics ?
nork: Ratnasingam, J., et al.
Argitaratua: (1999)

Riemannian walk for incremental learning: Understanding forgetting and intransigence
nork: Chaudhry, A, et al.
Argitaratua: (2018)

Early-career researchers: answering the most important scientific questions of our time
nork: Kelly, S
Argitaratua: (2021)

Structure and function of the centriole in animal cells: progress and questions.
nork: Lange, B, et al.
Argitaratua: (1996)

Graph inductive biases in transformers without message passing
nork: Ma, L, et al.
Argitaratua: (2023)

Graph inductive biases in transformers without message passing
nork: Ma, L, et al.
Argitaratua: (2023)

Deep learning for predicting COVID-19 malignant progression
nork: Fang, C, et al.
Argitaratua: (2021)

Multiple-instance learning with structured bag models
nork: Warrell, J, et al.
Argitaratua: (2011)

Learn what matters: cross-domain imitation learning with task-relevant embeddings
nork: Franzmeyer, T, et al.
Argitaratua: (2023)