> ML_LITERATURE_ATLAS_v1.0
Research Literature Atlas
253 qualified literature records from foundational statistical learning to frontier reasoning LLMs: verified DOIs, arXiv IDs, and original bilingual syntheses.
wav2vec 2.0: A Framework for Self-Supervised Learning of Speech Representations
The landmark Facebook AI paper establishing wav2vec 2.0, revolutionizing low-resource speech recognition through self-supervised pre-training.
Semi-Supervised Classification with Graph Convolutional Networks (GCN)
The foundational ICLR paper originating Graph Convolutional Networks (GCNs), creating modern graph representation learning across citation, fraud, and biological networks.
How Powerful are Graph Neural Networks? (Graph Isomorphism Network - GIN)
Foundational theoretical ICLR paper establishing the mathematical limits of graph neural networks and designing GIN for maximally discriminative graph classification.
Forecasting at Scale (Prophet)
Landmark Facebook research paper establishing Prophet, the most widely deployed automated forecasting tool in business operations, demand planning, and capacity engineering.
N-BEATS: Neural basis expansion analysis for interpretable time series forecasting
Landmark ICLR paper winning the M4 competition benchmarks, demonstrating that pure deep feedforward networks surpass complex statistical ensembles in time series.
A Time Series is Worth 64 Words: Long-term Forecasting with Transformers (PatchTST)
Influential ICLR paper establishing PatchTST, restoring Transformers as state-of-the-art contenders in long-horizon multivariate forecasting through patching and channel independence.
Are Transformers Effective for Time Series? (DLinear / NLinear)
Award-winning AAAI paper challenging the blind application of Transformers to time series, establishing DLinear as an indispensable minimalist benchmark.
Multilayer feedforward networks are universal approximators
Foundational mathematical paper proving that neural networks are universal function approximators, grounding connectionist modeling in functional analysis.
Bagging Predictors
Seminal paper establishing bootstrap aggregation, the core algorithmic mechanism behind Random Forests and bagging ensembles.
Induction of Decision Trees (ID3)
Foundational machine learning paper introducing information-theoretic decision tree learning, creating transparent and interpretable classification rules.
A Decision-Theoretic Generalization of On-Line Learning and an Application to Boosting (AdaBoost)
Gödel Prize-winning paper introducing AdaBoost, establishing the theoretical and mathematical foundations of ensemble boosting in machine learning.
Understanding the difficulty of training deep feedforward neural networks (Xavier Initialization)
Landmark AISTATS paper establishing Xavier weight initialization, resolving early gradient vanishing and enabling stable training of deep networks with sigmoid and tanh activations.
Delving Deep into Rectifiers: Surpassing Human-Level Performance on ImageNet Classification (PReLU / Kaiming Init)
Historic ICCV paper establishing Kaiming initialization and PReLU, standard across all modern deep neural networks with rectified activations (PyTorch default).
Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift
Monumental ICML paper establishing Batch Normalization, allowing networks to train 10x faster and eliminating hyper-sensitivity to parameter initialization.
Decoupled Weight Decay Regularization (AdamW)
The landmark ICLR paper introducing AdamW, the universal default optimizer for pretraining Transformers, BERT, GPT, and modern foundation models.
Adam: A Method for Stochastic Optimization
The most cited optimization paper in computer science, establishing Adam as the workhorse optimizer of deep learning.
Auto-Encoding Variational Bayes (VAE)
Monumental ICLR paper uniting variational inference with deep neural networks, forming the foundation of modern probabilistic deep learning.
Think you have Solved Question Answering? Try ARC, the AI2 Reasoning Challenge
The landmark Allen Institute for AI benchmark evaluating real-world scientific and qualitative reasoning, widely standard in open LLM leaderboards.