PyTorchKR Papers

최신 논문

총 215편
태그·기간 등 상세 필터

카테고리 탐색

로그인 후 이 태그 구독

arXiv:1807.03149

Learning models for visual 3D localization with implicit mapping

We consider learning based methods for visual localization that do not require the construction of explicit maps in the form of point clouds or voxels. The goal is to learn an implicit representation of the environment at a higher, more abstract level. We propose to use a generative approach based on Generative Query Networks (GQNs, Eslami et al. 2018), asking the following questions: 1) Can GQN capture more complex scenes than those it was originally demonstrated on? 2) Can GQN be used for localization in those scenes? To study this approach we consider procedurally generated Minecraft worlds, for which we can generate images of complex 3D scenes along with camera pose coordinates. We first show that GQNs, enhanced with a novel attention mechanism can capture the structure of 3D scenes in Minecraft, as evidenced by their samples. We then apply the models to the localization problem, comparing the results to a discriminative baseline, and comparing the ways each approach captures the task uncertainty.

arXiv:1406.1767

Changing the Environment Based on Empowerment as Intrinsic Motivation

One aspect of intelligence is the ability to restructure your own environment so that the world you live in becomes more beneficial to you. In this paper we investigate how the information-theoretic measure of agent empowerment can provide a task-independent, intrinsic motivation to restructure the world. We show how changes in embodiment and in the environment change the resulting behaviour of the agent and the artefacts left in the world. For this purpose, we introduce an approximation of the established empowerment formalism based on sparse sampling, which is simpler and significantly faster to compute for deterministic dynamics. Sparse sampling also introduces a degree of randomness into the decision making process, which turns out to beneficial for some cases. We then utilize the measure to generate agent behaviour for different agent embodiments in a Minecraft-inspired three dimensional block world. The paradigmatic results demonstrate that empowerment can be used as a suitable generic intrinsic motivation to not only generate actions in given static environments, as shown in the past, but also to modify existing environmental conditions. In doing so, the emerging strategies to modify an agent's environment turn out to be meaningful to the specific agent capabilities, i.e., de facto to its embodiment.

arXiv:1603.04119

Exploratory Gradient Boosting for Reinforcement Learning in Complex Domains

High-dimensional observations and complex real-world dynamics present major challenges in reinforcement learning for both function approximation and exploration. We address both of these challenges with two complementary techniques: First, we develop a gradient-boosting style, non-parametric function approximator for learning on $Q$-function residuals. And second, we propose an exploration strategy inspired by the principles of state abstraction and information acquisition under uncertainty. We demonstrate the empirical effectiveness of these techniques, first, as a preliminary check, on two standard tasks (Blackjack and $n$-Chain), and then on two much larger and more realistic tasks with high-dimensional observation spaces. Specifically, we introduce two benchmarks built within the game Minecraft where the observations are pixel arrays of the agent's visual field. A combination of our two algorithmic techniques performs competitively on the standard reinforcement-learning tasks while consistently and substantially outperforming baselines on the two tasks with high-dimensional observation spaces. The new function approximator, exploration strategy, and evaluation benchmarks are each of independent interest in the pursuit of reinforcement-learning methods that scale to real-world domains.

arXiv:1605.09128

Control of Memory, Active Perception, and Action in Minecraft

In this paper, we introduce a new set of reinforcement learning (RL) tasks in Minecraft (a flexible 3D world). We then use these tasks to systematically compare and contrast existing deep reinforcement learning (DRL) architectures with our new memory-based DRL architectures. These tasks are designed to emphasize, in a controllable manner, issues that pose challenges for RL methods including partial observability (due to first-person visual observations), delayed rewards, high-dimensional visual observations, and the need to use active perception in a correct manner so as to perform well in the tasks. While these tasks are conceptually simple to describe, by virtue of having all of these challenges simultaneously they are difficult for current DRL architectures. Additionally, we evaluate the generalization performance of the architectures on environments not used during training. The experimental results show that our new architectures generalize to unseen environments better than existing DRL architectures.

arXiv:1604.07255

A Deep Hierarchical Approach to Lifelong Learning in Minecraft

We propose a lifelong learning system that has the ability to reuse and transfer knowledge from one task to another while efficiently retaining the previously learned knowledge-base. Knowledge is transferred by learning reusable skills to solve tasks in Minecraft, a popular video game which is an unsolved and high-dimensional lifelong learning problem. These reusable skills, which we refer to as Deep Skill Networks, are then incorporated into our novel Hierarchical Deep Reinforcement Learning Network (H-DRLN) architecture using two techniques: (1) a deep skill array and (2) skill distillation, our novel variation of policy distillation (Rusu et. al. 2015) for learning skills. Skill distillation enables the HDRLN to efficiently retain knowledge and therefore scale in lifelong learning, by accumulating knowledge and encapsulating multiple reusable skills into a single distilled network. The H-DRLN exhibits superior performance and lower learning sample complexity compared to the regular Deep Q Network (Mnih et. al. 2015) in sub-domains of Minecraft.

arXiv:1705.00341

Deriving Quests from Open World Mechanics

Open world games present players with more freedom than games with linear progression structures. However, without clearly-defined objectives, they often leave players without a sense of purpose. Most of the time, quests and objectives are hand-authored and overlaid atop an open world's mechanics. But what if they could be generated organically from the gameplay itself? The goal of our project was to develop a model of the mechanics in Minecraft that could be used to determine the ideal placement of objectives in an open world setting. We formalized the game logic of Minecraft in terms of logical rules that can be manipulated in two ways: they may be executed to generate graphs representative of the player experience when playing an open world game with little developer direction; and they may be statically analyzed to determine dependency orderings, feedback loops, and bottlenecks. These analyses may then be used to place achievements on gameplay actions algorithmically.

arXiv:1709.03969

Explore, Exploit or Listen: Combining Human Feedback and Policy Model to Speed up Deep Reinforcement Learning in 3D Worlds

We describe a method to use discrete human feedback to enhance the performance of deep learning agents in virtual three-dimensional environments by extending deep-reinforcement learning to model the confidence and consistency of human feedback. This enables deep reinforcement learning algorithms to determine the most appropriate time to listen to the human feedback, exploit the current policy model, or explore the agent's environment. Managing the trade-off between these three strategies allows DRL agents to be robust to inconsistent or intermittent human feedback. Through experimentation using a synthetic oracle, we show that our technique improves the training speed and overall performance of deep reinforcement learning in navigating three-dimensional environments using Minecraft. We further show that our technique is robust to highly innacurate human feedback and can also operate when no human feedback is given.

arXiv:1710.11424

Regret Minimization for Partially Observable Deep Reinforcement Learning

Deep reinforcement learning algorithms that estimate state and state-action value functions have been shown to be effective in a variety of challenging domains, including learning control strategies from raw image pixels. However, algorithms that estimate state and state-action value functions typically assume a fully observed state and must compensate for partial observations by using finite length observation histories or recurrent networks. In this work, we propose a new deep reinforcement learning algorithm based on counterfactual regret minimization that iteratively updates an approximation to an advantage-like function and is robust to partially observed state. We demonstrate that this new algorithm can substantially outperform strong baseline methods on several partially observed reinforcement learning tasks: learning first-person 3D navigation in Doom and Minecraft, and acting in the presence of partially observed objects in Doom and Pong.

arXiv:1707.00183공식 코드 · Papers with Code

Teacher-Student Curriculum Learning

We propose Teacher-Student Curriculum Learning (TSCL), a framework for automatic curriculum learning, where the Student tries to learn a complex task and the Teacher automatically chooses subtasks from a given set for the Student to train on. We describe a family of Teacher algorithms that rely on the intuition that the Student should practice more those tasks on which it makes the fastest progress, i.e. where the slope of the learning curve is highest. In addition, the Teacher algorithms address the problem of forgetting by also choosing tasks where the Student's performance is getting worse. We demonstrate that TSCL matches or surpasses the results of carefully hand-crafted curricula in two tasks: addition of decimal numbers with LSTM and navigation in Minecraft. Using our automatically generated curriculum enabled to solve a Minecraft maze that could not be solved at all when training directly on solving the maze, and the learning was an order of magnitude faster than uniform sampling of subtasks.

arXiv:1712.07294

Hierarchical and Interpretable Skill Acquisition in Multi-task Reinforcement Learning

Learning policies for complex tasks that require multiple different skills is a major challenge in reinforcement learning (RL). It is also a requirement for its deployment in real-world scenarios. This paper proposes a novel framework for efficient multi-task reinforcement learning. Our framework trains agents to employ hierarchical policies that decide when to use a previously learned policy and when to learn a new skill. This enables agents to continually acquire new skills during different stages of training. Each learned task corresponds to a human language description. Because agents can only access previously learned skills through these descriptions, the agent can always provide a human-interpretable description of its choices. In order to help the agent learn the complex temporal dependencies necessary for the hierarchical policy, we provide it with a stochastic temporal grammar that modulates when to rely on previously learned skills and when to execute new skills. We validate our approach on Minecraft games designed to explicitly test the ability to reuse previously learned skills while simultaneously learning new skills.

arXiv:1706.05170

Interactive 3D Modeling with a Generative Adversarial Network

This paper proposes the idea of using a generative adversarial network (GAN) to assist a novice user in designing real-world shapes with a simple interface. The user edits a voxel grid with a painting interface (like Minecraft). Yet, at any time, he/she can execute a SNAP command, which projects the current voxel grid onto a latent shape manifold with a learned projection operator and then generates a similar, but more realistic, shape using a learned generator network. Then the user can edit the resulting shape and snap again until he/she is satisfied with the result. The main advantage of this approach is that the projection and generation operators assist novice users to create 3D models characteristic of a background distribution of object shapes, but without having to specify all the details. The core new research idea is to use a GAN to support this application. 3D GANs have previously been used for shape generation, interpolation, and completion, but never for interactive modeling. The new challenge for this application is to learn a projection operator that takes an arbitrary 3D voxel model and produces a latent vector on the shape manifold from which a similar and realistic shape can be generated. We develop algorithms for this and other steps of the SNAP processing pipeline and integrate them into a simple modeling tool. Experiments with these algorithms and tool suggest that GANs provide a promising approach to computer-assisted interactive modeling.

arXiv:1803.08456

Deep Reinforcement Learning with Model Learning and Monte Carlo Tree Search in Minecraft

Deep reinforcement learning has been successfully applied to several visual-input tasks using model-free methods. In this paper, we propose a model-based approach that combines learning a DNN-based transition model with Monte Carlo tree search to solve a block-placing task in Minecraft. Our learned transition model predicts the next frame and the rewards one step ahead given the last four frames of the agent's first-person-view image and the current action. Then a Monte Carlo tree search algorithm uses this model to plan the best sequence of actions for the agent to perform. On the proposed task in Minecraft, our model-based approach reaches the performance comparable to the Deep Q-Network's, but learns faster and, thus, is more training sample efficient.

arXiv:1803.09866

Accelerating Empowerment Computation with UCT Tree Search

Models of intrinsic motivation present an important means to produce sensible behaviour in the absence of extrinsic rewards. Applications in video games are varied, and range from intrinsically motivated general game-playing agents to non-player characters such as companions and enemies. The information-theoretic quantity of Empowerment is a particularly promising candidate motivation to produce believable, generic and robust behaviour. However, while it can be used in the absence of external reward functions that would need to be crafted and learned, empowerment is computationally expensive. In this paper, we propose a modified UCT tree search method to mitigate empowerment's computational complexity in discrete and deterministic scenarios. We demonstrate how to modify a Monte-Carlo Search Tree with UCT to realise empowerment maximisation, and discuss three additional modifications that facilitate better sampling. We evaluate the approach both quantitatively, by analysing how close our approach gets to the baseline of exhaustive empowerment computation with varying amounts of computational resources, and qualitatively, by analysing the resulting behaviour in a Minecraft-like scenario.

arXiv:1803.09853

Generative Design in Minecraft (GDMC), Settlement Generation Competition

This paper introduces the settlement generation competition for Minecraft, the first part of the Generative Design in Minecraft challenge. The settlement generation competition is about creating Artificial Intelligence (AI) agents that can produce functional, aesthetically appealing and believable settlements adapted to a given Minecraft map - ideally at a level that can compete with human created designs. The aim of the competition is to advance procedural content generation for games, especially in overcoming the challenges of adaptive and holistic PCG. The paper introduces the technical details of the challenge, but mostly focuses on what challenges this competition provides and why they are scientifically relevant.

arXiv:1805.07008

Hierarchical Reinforcement Learning with Deep Nested Agents

Deep hierarchical reinforcement learning has gained a lot of attention in recent years due to its ability to produce state-of-the-art results in challenging environments where non-hierarchical frameworks fail to learn useful policies. However, as problem domains become more complex, deep hierarchical reinforcement learning can become inefficient, leading to longer convergence times and poor performance. We introduce the Deep Nested Agent framework, which is a variant of deep hierarchical reinforcement learning where information from the main agent is propagated to the low level $nested$ agent by incorporating this information into the nested agent's state. We demonstrate the effectiveness and performance of the Deep Nested Agent framework by applying it to three scenarios in Minecraft with comparisons to a deep non-hierarchical single agent framework, as well as, a deep hierarchical framework.