Deep RL algorithm learns human-level policies on nearly all Atari games.
problem Consistent performance on diverse Atari games.
method Addressing three key challenges: diverse reward distributions, long-term reasoning, and efficient exploration.
result Exceeds human performance on 40 out of 42 Atari games.
Agents learn to play a first-person multiplayer game at human level performance.
problem Training AI agents for complex, multi-agent, real-time environments.
method Population-based deep reinforcement learning with concurrent training of multiple agents.
result Achieved human-level performance in a first-person multiplayer game.
DreamerV2 learns Atari game behaviors from a world model, achieving human-level performance.
problem Learning complex behaviors in Atari games from limited data.
method DreamerV2 uses a world model with discrete representations to predict behaviors in a compact latent space.
result Achieves human-level performance on 55 Atari tasks.
Study finds resolution impacts human classification performance in MNIST data.
problem Understanding factors affecting human classification performance in machine learning.
method Empirical study of MNIST data at various resolutions.
result Derived a quantitative relationship between resolution and human classification performance.
Face verification remains a challenging problem in very complex conditions with large variations such as pose, illumination, expression, and occlusions. This problem is exacerbated when we rely unrealistically on a single training data source, which is often insufficient to cover the intrinsically complex face variatio…
Paper presents a DRL framework for detecting and anticipating financial crises.
problem Detecting and adapting to financial crises using deep reinforcement learning.
method Two sub-networks, one for past performances and standard deviations, the other for contextual features. Adversarial training for robustness.
result Framework substantially outperforms traditional methods in detecting and anticipating crises.
Study benchmarks automated sleep staging against human scorers, achieving human-level performance.
problem Lack of standardized comparison between human and automated sleep staging.
method Developed multi-scored datasets and a framework to compare multiple human scorers' consensus.
result Many automated methods can match human scorers' performance, with SimpleSleepNet achieving high F1 scores.
Cyclic Boosting offers detailed prediction understanding for machine learning models.
problem Complex machine learning models are often black boxes, making individual predictions hard to understand.
method Cyclic Boosting is a novel machine learning algorithm that provides detailed understanding of predictions.
result Cyclic Boosting enables detailed understanding of how predictions are made, even for complex models.
Neural networks have been successfully applied in applications with a large amount of labeled data. However, the task of rapid generalization on new concepts with small training data while preserving performances on previously learned ones still presents a significant challenge to neural network models. In this work, w…
Improved deep Q-learning with SVRG reduces variance and stabilizes training.
problem Excessive variance in gradient estimation hinders deep Q-learning performance.
method Utilized stochastic variance reduced gradient (SVRG) techniques.
result Significantly improved performance on 18 out of 20 Atari games compared to baseline methods.
Animal vision surpasses human tech, but we can't replicate their abilities.
problem Can we replicate animal vision capabilities?
method Comparing animal vision systems to human technological achievements.
result Human tech can't match animal vision capabilities.
State of the art deep reinforcement learning algorithms take many millions of interactions to attain human-level performance. Humans, on the other hand, can very quickly exploit highly rewarding nuances of an environment upon first discovery. In the brain, such rapid learning is thought to depend on the hippocampus and…
EBU improves deep RL with episodic backward updates, achieving human-level performance with less data.
problem Efficiently learn from sparse and delayed rewards in reinforcement learning.
method Episodic Backward Update (EBU) algorithm that recursively propagates value updates.
result EBU achieves human-level performance in Atari games with significantly fewer samples.
Machine learning lacks causal models, hindering strong AI.
problem Current machine learning systems lack a model of reality.
method Demonstrates seven tasks beyond current ML capabilities using causal inference.
result Machine learning needs causal models for strong AI.
The paper improves NLI models' robustness by adding external knowledge to the attention mechanism.
problem NLI models' performance drops significantly on simple adversarial examples.
method Proposes a method to enhance NLI models' robustness by incorporating external knowledge into the attention mechanism.
result The method significantly improves the robustness of NLI models, achieving human-level performance on adversarial data.
New research shows label refinement and weak training have limitations for aligning LLMs.
problem Limitations of refinement methods for aligning large language models.
method Analyzed probabilistic assumptions and alternative approaches to label refinement and weak training.
result Label refinement and weak training suffer from irreducible error, leaving a performance gap.
Paper explores trading strategies using LSTM and reinforcement learning.
problem Achieving human-level trading performance with AI.
method Reinforcement Learning and LSTM neural networks.
result Robust trading system verified on GBPUSD trading.
AFLite filters dataset biases to improve model generalization.
problem Overfitting to dataset biases degrades model performance on out-of-distribution samples.
method AFLite adversarially filters dataset biases, reducing their impact on model performance.
result Filtered datasets yield better generalization to out-of-distribution tasks.
Improved deep learning for action recognition using multi-kernel SVM and deep neural networks.
problem Challenges in video understanding, especially action recognition, despite deep neural networks' success in image understanding.
method Combining multi-kernel SVM with a multi-stream deep convolutional neural network, including hand-crafted features.
result Achieved close to state-of-the-art performance on the HMDB-51 dataset.
A method to train deep neural networks on noisy labeled data.
problem Training deep neural networks on noisy labeled data causes performance degradation.
method A noise-tolerant training algorithm that simulates actual training with synthetic noisy labels.
result The proposed method outperforms state-of-the-art baselines on noisy CIFAR-10 and Clothing1M datasets.
Paper proposes an alternative to anchor points for learning with noisy labels.
problem Learning with noisy labels is challenging due to inaccurate labels.
method Estimates transition matrix using clusterability condition and noisy labels.
result Estimation of transition matrix is more accurate and efficient than anchor points.
Adversarial examples and noisy images share a common cause.
problem Improving machine learning models' robustness to adversarial attacks and random noise.
method Empirical and theoretical analysis of adversarial examples and corrupted images.
result Adversarial robustness and corruption robustness are closely related.
Deep learning improves reinforcement learning in video games.
problem Challenges in reinforcement learning with high-dimensional inputs.
method Used deep Q-network and batch normalization to improve reinforcement learning.
result Some agents learned to play T-rex Runner better than human experts.
Generative grading improves automated feedback for structured problems.
problem Difficulty in providing high-quality feedback on structured assignments.
method Generative descriptions of student cognition, probabilistic programs, and learning to infer feedback.
result Achieved near-human accuracy in grading and feedback across diverse domains.
RS-Rainbow learns to interpret Atari agent decisions.
problem Difficult to understand how Atari agents make decisions.
method End-to-end trainable network with attention module.
result Improves model interpretability and performance.
Generative classifiers show surprising human-like performance.
problem Comparing generative and discriminative models for object recognition.
method Built on recent advances in generative modeling to create classifiers and compared them to discriminative models.
result Generative classifiers outperform discriminative models in several key areas, including shape bias and out-of-distribution accuracy.
A simple method to prevent forgetting in neural networks.
problem Catastrophic forgetting in neural networks.
method Weight friction, inspired by neurology and physics, modifies the gradient descent update rule.
result Weight friction enables continual learning with comparable performance and efficiency.
Report on challenges and approaches for Multi-Agent RL.
problem Challenges in Multi-Agent RL for cooperative and competitive environments.
method Decentralized Actor, Centralized Critic approach based on Decentralized Partially Observable MDPs.
result Advances in Multi-Agent RL for mixed cooperative and competitive environments.
Compact convolution enhances CNN robustness against adversarial attacks.
problem Improving network robustness against adversarial attacks.
method Learning compact features using L2-Softmax Loss and compact convolution.
result Compact Convolutional Networks (CCNs) neutralize multiple types of adversarial attacks.
BC learning improves deep sound recognition performance.
problem Improving deep sound recognition using novel training data.
method BC learning: mixing sounds from different classes to generate between-class sounds and train models to recognize these.
result BC learning improves performance on various sound recognition networks, surpassing human level.
Efficient Q-learning for metric spaces without planning oracles.
problem Efficient reinforcement learning in continuous state-action spaces.
method Model-free Q-learning adapted for metric spaces without a planning oracle.
result Sample efficiency in continuous state-action spaces.
Survey on DRL for clinical decision support.
problem Improving clinical decision-making through AI.
method Deep reinforcement learning with DNN.
result Survey and comparison of DRL algorithms in clinical applications.
Efficient facial feature learning with shared representations reduces redundancy and improves accuracy.
problem Redundancy and high computational load in training deep ensemble models.
method Wide Ensemble-based Convolutional Neural Networks (ESRs) with varying branching levels.
result ESRs reduce residual generalization error and outperform state-of-the-art methods on facial expression recognition.
New method evaluates language model forecasters by checking consistency of predictions.
problem Evaluating the performance of language model forecasters is difficult due to lack of ground truth.
method Developed a consistency check framework based on arbitrage to evaluate forecasters.
result Consistency metrics correlate with ground truth performance of LLM forecasters.
Flow IV uses IVs to infer counterfactuals in complex models.
problem Identifying causal effects and counterfactual reasoning in nonseparable outcome models.
method Utilizes instrumental variables and normalizing flows to estimate and infer counterfactual outcomes.
result Identifies a method to make causal inferences from observed data in nonseparable models.
This study examines how hidden layers affect CNN performance on handwritten digit recognition.
problem Impact of hidden layers on CNN performance in handwritten digit recognition.
method Applied CNN with varying hidden layers on MNIST dataset, trained with stochastic gradient and backpropagation, tested with feedforward.
result Variations in accuracies for different hidden layers and epochs.
Paper proposes MAG to fine-tune BERT and XLNet for multimodal sentiment analysis.
problem Fine-tuning pre-trained models for multimodal language applications is challenging.
method Integrates vision and acoustic modalities into BERT and XLNet through Multimodal Adaptation Gate (MAG).
result Significant improvement in multimodal sentiment analysis performance over previous methods.
COMET learns concepts for few-shot learning, improving performance.
problem Few-shot learning challenges in machine learning.
method Meta-learning with human-interpretable concept dimensions.
result Significant improvement in 1-shot learning tasks.
Deep learning models generate languages that lack abstract reasoning.
problem Lack of abstract reasoning in deep learning-generated languages.
method Analyzed emergent language from two multi-agent games with compositional measures.
result Deep learning solutions often fail to generalize to out-of-training examples.
Proposes MANet for efficient DRL with less experience samples.
problem Inefficient DRL due to lack of spatial abstraction and attention.
method Divides input into partial states, parallel attention layers attend to relevant partial states, estimates state-action values.
result Significantly less experience samples for high performance.
A model disentangles features to detect adversarial inputs.
problem Detecting and defending against adversarial attacks on neural networks.
method Proposes a minimax game formulation using variational autoencoders to separate robust and vulnerable features.
result Adversarial inputs cannot bypass the detector without semantic change, indicating successful detection.
Agent learns to play hard games by watching YouTube videos.
problem Sparse rewards in reinforcement learning environments.
method Self-supervised video mapping, YouTube video embedding, imitation reward function.
result Agent achieves human-level performance on hard games.
Novel meta-RL strategy improves efficiency in learning novel tasks.
problem Efficiency in learning novel tasks using deep RL.
method Decomposes meta-RL into task-exploration, task-inference, and task-fulfillment; uses deep networks and a task encoder.
result Improves sample efficiency and mitigates meta-overfitting.
Paper analyzes self-supervised learning using causal methods and proposes a new objective.
problem Lack of theoretical understanding of self-supervised learning success.
method Uses a causal framework to enforce invariance constraints on proxy classifiers.
result ReLIC objective improves generalization guarantees and outperforms existing methods.
This paper proposes an ADMM-based method for progressive weight pruning of deep neural networks.
problem Large model size of deep neural networks hinder their applications on edge devices.
method Progressive weight pruning using ADMM for non-convex optimization problems.
result Achieves up to 34 times pruning rate for ImageNet and 167 times for MNIST datasets.
This research tackles image classification with noise, proposing committees of CNNs.
problem Image classification with concurrent feature and label noise.
method Committees of Convolutional Neural Networks (CNNs) for MNIST, CIFAR-10, and CIFAR-100 datasets.
result Committees outperform single models in noisy conditions, especially on difficult datasets.
VISR learns controllable features for fast task inference.
problem Generalizing behaviors beyond explicitly learned set for subsequent tasks.
method Combines Successor Features and Variational Inference.
result Achieves human-level performance on 14 Atari games.
Randomized smoothing reduces accuracy in ML models, especially at higher noise levels.
problem Adversarial attacks on ML models, especially randomized smoothing's accuracy drop.
method Theoretical and empirical analysis of randomized smoothing's effect on feasible hypotheses space.
result For some noise levels, randomized smoothing shrinks the set of feasible hypotheses, leading to accuracy drops.