932 Hits in 8.7 sec

Deep(er) learning

Shyam Srinivasan, Ralph J. Greenspan, Charles F. Stevens, Dhruv Grover
2018 Journal of Neuroscience  
With judicious use, deep learning can become yet another effective tool to understand how and why brains are designed the way they are.  ...  In this viewpoint, we advocate that deep learning can be further enhanced by incorporating and tightly integrating five fundamental principles of neural circuit design and function: optimizing the system  ...  Using a deep neural network to guide its search, AlphaGO chose the move that was most successful in simulation.  ... 
doi:10.1523/jneurosci.0153-18.2018 pmid:30006366 pmcid:PMC6596136 fatcat:2guzm6y2a5fcblt4773nkg7f24

Meta-Learning in Neural Networks: A Survey [article]

Timothy Hospedales, Antreas Antoniou, Paul Micaelli, Amos Storkey
2020 arXiv   pre-print
We survey promising applications and successes of meta-learning such as few-shot learning and reinforcement learning.  ...  We first discuss definitions of meta-learning and position it with respect to related fields, such as transfer learning and hyperparameter optimization.  ...  to -the conventional sparse reward objective.  ... 
arXiv:2004.05439v2 fatcat:3r23tsxxkfbgzamow5miglkrye

A Survey of Deep Learning Applications to Autonomous Vehicle Control [article]

Sampo Kuutti, Richard Bowden, Yaochu Jin, Phil Barber, Saber Fallah
2019 arXiv   pre-print
For these reasons, the use of deep learning for vehicle control is becoming increasingly popular.  ...  and object detection.  ...  This is known as the black box problem.  ... 
arXiv:1912.10773v1 fatcat:vtmdnxgt7zdadnlyn7cfyvldai

Deep Reinforcement Learning for Complex Manipulation Tasks with Sparse Feedback [article]

Binyamin Manela
2020 arXiv   pre-print
Learning optimal policies from sparse feedback is a known challenge in reinforcement learning.  ...  To test our algorithms, we built three challenging manipulation environments with sparse reward functions. Each environment has three levels of complexity.  ...  Acknowledgements This research was supported in part by the Helmsley Charitable Trust through the Agricultural, Biological and Cognitive Robotics Initiative and by the Marcus Endowment Fund both at the  ... 
arXiv:2001.03877v1 fatcat:7ucagopk6vbxjlssxnqimfllum

DANCin SEQ2SEQ: Fooling Text Classifiers with Adversarial Text Example Generation [article]

Catherine Wong
2017 arXiv   pre-print
We recast adversarial text example generation as a reinforcement learning problem, and demonstrate that our algorithm offers preliminary but promising steps towards generating semantically meaningful adversarial  ...  In this work, we introduce DANCin SEQ2SEQ, a GAN-inspired algorithm for adversarial text example generation targeting largely black-box text classifiers.  ...  Additional thanks to Animesh Garg and Charles Lu for guiding discussion on the theme of "how can I fool classifiers with reinforcement learning", and finally, to George Fei, for encouraging all kinds of  ... 
arXiv:1712.05419v1 fatcat:ccrkfg4nargw3hctkdi6iispym

Meta-Learning in Neural Networks: A Survey

Timothy M Hospedales, Antreas Antoniou, Paul Micaelli, Amos J. Storkey
2021 IEEE Transactions on Pattern Analysis and Machine Intelligence  
We survey promising applications and successes of meta-learning including few-shot learning, reinforcement learning and architecture search.  ...  We first discuss definitions of meta-learning and position it with respect to related fields, such as transfer learning, multi-task learning, and hyperparameter optimization.  ...  to -the conventional sparse reward objective.  ... 
doi:10.1109/tpami.2021.3079209 pmid:33974543 fatcat:wkzeodki4fbcnjlcczn4mr6kry

Flexible and Efficient Long-Range Planning Through Curious Exploration [article]

Aidan Curtis, Minjian Xin, Dilip Arumugam, Kevin Feigelis, Daniel Yamins
2020 arXiv   pre-print
In contrast, deep reinforcement learning (DRL) methods use flexible neural-network-based function approximators to discover policies that generalize naturally to unseen circumstances.  ...  However, DRL methods struggle to handle the very sparse reward landscapes inherent to long-range multi-step planning situations.  ...  and functioning in environments where the reward it not sparse, but fails to learn in sparse-reward environments even when the reward includes intrinsic curiosity. random point on the no-reward trajectory  ... 
arXiv:2004.10876v2 fatcat:4ioyp3qubnhvtg7txjjkzuixp4

Deep Learning and Reinforcement Learning for Autonomous Unmanned Aerial Systems: Roadmap for Theory to Deployment [article]

Jithin Jagannath, Anu Jagannath, Sean Furman, Tyler Gwin
2020 arXiv   pre-print
We then provide an overview of some of the key deep learning and reinforcement learning techniques discussed throughout this chapter.  ...  Therefore, in this chapter, we discuss how some of the advances in machine learning, specifically deep learning and reinforcement learning can be leveraged to develop next-generation autonomous UAS.  ...  Reinforcement Learning for Autonomous UAS Deep Learning and Reinforcement Learning for Autonomous UAS Deep Learning and Reinforcement Learning for Autonomous UAS Deep Learning and Reinforcement  ... 
arXiv:2009.03349v2 fatcat:5ylreoukrfcrtorzzp44mntjum

End-to-End Neuro-Symbolic Architecture for Image-to-Image Reasoning Tasks [article]

Ananye Agarwal, Pradeep Shenoy, Mausam
2021 arXiv   pre-print
Neural models and symbolic algorithms have recently been combined for tasks requiring both perception and reasoning.  ...  Experiments show high accuracy with significantly less data compared to purely neural approaches.  ...  difficulty of learning in large, sparse-reward spaces, especially when reward distributions are non-smooth.  ... 
arXiv:2106.03121v1 fatcat:kzu4pmlnurah5f2mstdkdnayby

Distributed Policy Iteration for Scalable Approximation of Cooperative Multi-Agent Policies [article]

Thomy Phan, Kyrill Schmid, Lenz Belzner, Thomas Gabor, Sebastian Feld, Claudia Linnhoff-Popien
2019 arXiv   pre-print
STEP combines decentralized multi-agent planning with centralized learning, only requiring a generative model for distributed black box optimization.  ...  We experimentally evaluate STEP in two challenging and stochastic domains with large state and joint action spaces and show that STEP is able to learn stronger policies than standard multi-agent reinforcement  ...  RELATED WORK Policy Iteration with Deep Learning and Tree Search.  ... 
arXiv:1901.08761v1 fatcat:d2t6m3dxyvfvlfz5hs4texjvga

Knowledge-Integrated Informed AI for National Security [article]

Anu K. Myne, Kevin J. Leahy, Ryan J. Soklaski
2022 arXiv   pre-print
Specifically, we review illuminating examples of knowledge integrated in deep learning and reinforcement learning pipelines, taking note of the performance gains they provide.  ...  While AI technology has and continues to become increasingly mainstream with impact across domains and industries, it's not without several drawbacks, weaknesses, and potential to cause undesired effects  ...  black-box, unaware neural networks.  ... 
arXiv:2202.03188v1 fatcat:sabdacy56ngizpo54ukxtg3o7y

Improving Exploration in Evolution Strategies for Deep Reinforcement Learning via a Population of Novelty-Seeking Agents [article]

Edoardo Conti, Vashisht Madhavan, Felipe Petroski Such, Joel Lehman, Kenneth O. Stanley, Jeff Clune
2018 arXiv   pre-print
Evolution strategies (ES) are a family of black-box optimization algorithms able to train deep neural networks roughly as well as Q-learning and policy gradient methods on challenging deep reinforcement  ...  However, many RL problems require directed exploration because they have reward functions that are sparse or deceptive (i.e. contain local optima), and it is unknown how to encourage such exploration with  ...  We also thank Justin Pinkul, Mike Deats, Cody Yancey, Joel Snow, Leon Rosenshein and the entire OpusStack Team inside Uber for providing our computing platform and for technical support.  ... 
arXiv:1712.06560v3 fatcat:3qgjexfqvbfyfeitkz3lnebpzu

A Survey of Machine Learning Applied to Computer Architecture Design [article]

Drew D. Penney, Lizhong Chen
2019 arXiv   pre-print
Machine learning has enabled significant benefits in diverse fields, but, with a few exceptions, has had limited impact on computer architecture.  ...  Taken together, these strategies and techniques present a promising future for increasingly automated architectural design.  ...  Reinforcement Learning: In reinforcement learning, an agent is sequentially provided with input based on an environment state and learns to perform actions that optimize a reward.  ... 
arXiv:1909.12373v1 fatcat:o4nscgkjfbes7kqwmtjvvgl3oa

DNN-Opt: An RL Inspired Optimization for Analog Circuit Sizing using Deep Neural Networks [article]

Ahmet F. Budak, Prateek Bhansali, Bo Liu, Nan Sun, David Z. Pan, Chandramouli V. Kashyap
2021 arXiv   pre-print
This paper presents DNN-Opt, a Reinforcement Learning (RL) inspired Deep Neural Network (DNN) based black-box optimization framework for analog circuit sizing.  ...  Our method shows 5--30x sample efficiency compared to other black-box optimization methods both on small building blocks and on large industrial circuits with better performance metrics.  ...  ACKNOWLEDGEMENT This work is supported in part by NSF under Grant No. 1704758.  ... 
arXiv:2110.00211v1 fatcat:ntui74flwzbipn6yntf64n4xu4

Solving Compositional Reinforcement Learning Problems via Task Reduction [article]

Yunfei Li, Yilin Wu, Huazhe Xu, Xiaolong Wang, Yi Wu
2021 arXiv   pre-print
Experiment results show that SIR can significantly accelerate and improve learning on a variety of challenging sparse-reward continuous-control problems with compositional structures.  ...  We propose a novel learning paradigm, Self-Imitation via Reduction (SIR), for solving compositional reinforcement learning problems. SIR is based on two core ideas: task reduction and self-imitation.  ...  Soft actor-critic: Off-policy maximum entropy deep reinforcement learning with a stochastic actor.  ... 
arXiv:2103.07607v2 fatcat:zspegy4xtjgvpimonjylblodme
« Previous Showing results 1 — 15 out of 932 results