Add to ORKGDeep learning methods have recently started dominating the machine learning world as they offer state-of-the-art performance in many applications. Unfortunately, the learning and evaluation of deep models is time-consuming, sometimes prohibitive, on resource-constrained devices such as phones and mobile hardware. Conditional computation attempts to alleviate this problem by selectively computing only parts of some model at a time. This thesis investigates the use of reinforcement learning algorithms as a tool to learn the computation selection strategies inside deep neural network models. Efficient ways of structuring these deep models are proposed, as well as parameterizations for activation-dependent computation policies. A learning schem...
Using deep neural networks for reinforcement learning has proven very successful, as demonstrated by...
In recent years there has been a growing attention from the world of research and companies in the f...
We present a data-efficient framework for solving sequential decision-making problems which exploits...
Summarization: Many computational problems can be solved by multiple algorithms, with different algo...
This thesis is mostly focused on reinforcement learning, which is viewed as an optimization problem:...
This thesis is mostly focused on reinforcement learning, which is viewed as an optimization problem:...
Recently, deep learning models such as convolutional and recurrent neural networks have displaced st...
Reinforcement learning (RL) is a broad family of algorithms for training autonomous agents to collec...
Reinforcement learning (RL) is a broad family of algorithms for training autonomous agents to collec...
Despite the success of reinforcement learning methods in various simulated robotic applications, end...
Using deep neural networks for reinforcement learning has proven very successful, as demonstrated by...
Using deep neural networks for reinforcement learning has proven very successful, as demonstrated by...
Recent advancements in deep reinforcement learning for real control tasks have received interest fro...
Many computational problems can be solved by multiple algorithms, with different algorithms fastest ...
In this paper, the Reinforcement Learning problem is formulated equivalently to a Markov Decision Pr...
Using deep neural networks for reinforcement learning has proven very successful, as demonstrated by...
In recent years there has been a growing attention from the world of research and companies in the f...
We present a data-efficient framework for solving sequential decision-making problems which exploits...
Summarization: Many computational problems can be solved by multiple algorithms, with different algo...
This thesis is mostly focused on reinforcement learning, which is viewed as an optimization problem:...
This thesis is mostly focused on reinforcement learning, which is viewed as an optimization problem:...
Recently, deep learning models such as convolutional and recurrent neural networks have displaced st...
Reinforcement learning (RL) is a broad family of algorithms for training autonomous agents to collec...
Reinforcement learning (RL) is a broad family of algorithms for training autonomous agents to collec...
Despite the success of reinforcement learning methods in various simulated robotic applications, end...
Using deep neural networks for reinforcement learning has proven very successful, as demonstrated by...
Using deep neural networks for reinforcement learning has proven very successful, as demonstrated by...
Recent advancements in deep reinforcement learning for real control tasks have received interest fro...
Many computational problems can be solved by multiple algorithms, with different algorithms fastest ...
In this paper, the Reinforcement Learning problem is formulated equivalently to a Markov Decision Pr...
Using deep neural networks for reinforcement learning has proven very successful, as demonstrated by...
In recent years there has been a growing attention from the world of research and companies in the f...
We present a data-efficient framework for solving sequential decision-making problems which exploits...