Mae
PulseAugur coverage of Mae — every cluster mentioning Mae across labs, papers, and developer communities, ranked by signal.
14 day(s) with sentiment data
-
Loss functions explained: MSE, MAE, Huber, and cross-entropy
The article explains the dual role of loss functions in machine learning: quantifying errors and guiding model training through their derivatives. It details how Mean Squared Error (MSE) converges to the mean and Mean A…
-
New Curia-MAE method enhances 3D medical image segmentation
Researchers have developed Curia-MAE, a new pre-training method for 3D medical image segmentation that aims to improve upon existing foundation models. This method incorporates a robust reconstruction objective, a featu…
-
MoCA framework enhances multi-modal wearable data analysis
Researchers have introduced MoCA, a novel self-supervised learning framework designed for analyzing multi-modal data from wearable devices. This framework utilizes a transformer architecture combined with masked autoenc…
-
New SLiM framework unifies skeleton learning with compact tokens
Researchers have developed SLiM, a novel framework for skeleton representation learning that unifies masked feature prediction and contrastive learning. This approach aims to overcome limitations in current methods by f…
-
Image generation difficulty depends on target representation, study finds
A new research paper explores how different target representations impact image generation difficulty. The study compared raw pixels, SD-VAE latents, DINOv2, and MAE features within a unified masked autoregressive model…
-
Foundation model pretraining strategies impact retinal imaging transferability
A new arXiv paper explores how different pretraining strategies for foundation models impact their effectiveness when transferred to ultra-widefield retinal imaging tasks. Researchers compared Vision Transformer encoder…
-
New framework audits volatility forecasts across market regimes
This paper introduces a novel framework for auditing volatility forecasts, moving beyond aggregate accuracy metrics like RMSE and MAE. The proposed method identifies latent market regimes and evaluates forecast reliabil…
-
New method aligns AI self-supervised learning with scientific imaging physics
Researchers have developed a new method for designing data augmentations in self-supervised learning (SSL) specifically for scientific imaging. This approach, termed physics-aligned augmentation, considers the unique sy…
-
New XAI-Enhanced Quantum Adversarial Networks Developed for Galaxy Modeling
Researchers have developed a novel quantum adversarial framework that combines a hybrid quantum neural network (QNN) with classical deep learning layers. This approach integrates an evaluator model using Local Interpret…
-
TimesFM 2.5 enhances time-series forecasting with new features
TimesFM 2.5, a time-series forecasting model, has been updated to include advanced features for end-to-end workflow development. The new version supports backtesting, covariate integration, anomaly detection, and scalab…
-
OpenAI releases Codex Security CLI; Open Dreamer launches; Grok 4.5 gets India pricing
OpenAI has released Codex Security CLI, a local AI agent designed to automatically detect and fix vulnerabilities within code repositories. Separately, independent researchers have launched Open Dreamer, an open-source …
-
New Bootleg method enhances self-supervised learning for AI models
Researchers have developed a new self-supervised learning method called Bootleg, which aims to combine the stability of generative approaches with the efficiency of predictive methods. Bootleg trains a model to predict …
-
JEPA predictors prove portable for occluded feature completion
Researchers have demonstrated that the predictor component of Joint-Embedding Predictive Architectures (JEPAs), typically discarded after training, can be repurposed as a transferable operator for occluded feature compl…
-
New framework improves retail demand forecasting with adaptive correction
Researchers have developed a new framework called Predict-then-Correct (PtC) to improve retail demand forecasting, particularly for situations with rapidly changing demand and limited early data. This framework combines…
-
New Neural Process Model Enhances Residential Load Forecasting
Researchers have developed a new behavior-conditioned Attentive Neural Process (ANP) framework for short-term load forecasting in residential settings. This model embeds inferred behavioral structure directly into the f…
-
AI scientist workflows show transferable discoveries in materials science · 2 sources tracked
Researchers have developed auditable AI-scientist workflows designed to ensure that AI-driven discoveries in materials science are robust and transferable. The study involved seven distinct search processes that evaluat…
-
New loss function APAL improves time-series forecasting for peak prediction
Researchers have developed a new loss function called Asymmetric Peak-Aware Loss (APAL) designed to improve time-series forecasting, particularly for applications where under-prediction carries higher risks than over-pr…
-
New research evaluates vision models' human-like color perception
A new research paper explores how well vision models understand color representation compared to humans. The study introduces a framework to evaluate color grounding based on human perceptual data, assessing category bo…
-
Research questions Transformer necessity for traffic forecasting
A new research paper questions the necessity of Transformers for extracting global spatial information in traffic forecasting. The study proposes an alternative approach using a simple global aggregation operator, which…
-
New method probes geospatial SSL representations using environmental signals
Researchers have developed a new method to evaluate self-supervised learning (SSL) representations in geospatial satellite imagery. Instead of relying solely on downstream tasks, this approach probes the representations…