# David Ha research scientist at google brain 🧠 in tokyo🗼 ## Featured Co-authors - [Kyunghyun Cho](/content/profile/kyunghyun-cho/index.html) 218 publications - [Honglak Lee](/content/profile/honglak-lee/index.html) 129 publications - [Julian Togelius](/content/profile/julian-togelius/index.html) 119 publications - [Joan Bruna](/content/profile/joan-bruna/index.html) 96 publications - [Jürgen Schmidhuber](/content/profile/jurgen-schmidhuber/index.html) 84 publications - [Marco Cuturi](/content/profile/marco-cuturi/index.html) 70 publications - [Sebastian Risi](/content/profile/sebastian-risi/index.html) 57 publications - [Timothy Lillicrap](/content/profile/timothy-lillicrap/index.html) 52 publications - [Jakob Foerster](/content/profile/jakob-foerster/index.html) 51 publications - [Charles Blundell](/content/profile/charles-blundell/index.html) 45 publications - [Alex Lamb](/content/profile/alex-lamb/index.html) 37 publications ### [Learning to Generalize with Object-centric Agents in the Open World Survival Game Crafter](/content/publication/learning-to-generalize-with-object-centric-agents-in-the-open-world-survival-game-crafter/index.html) Reinforcement learning agents must generalize beyond their training. ### [Simultaneous Multiple-Prompt Guided Generation Using Differentiable Optimal Transport](/content/publication/simultaneous-multiple-prompt-guided-generation-using-differentiable-optimal-transport/index.html) Recent advances in deep learning, such as powerful generative models and... ### [Evolving Modular Soft Robots without Explicit Inter-Module Communication using Local Self-Attention](/content/publication/evolving-modular-soft-robots-without-explicit-inter-module-communication-using-local-self-attention/index.html) Modularity in robotics holds great potential. ### [EvoJAX: Hardware-Accelerated Neuroevolution](/content/publication/evojax-hardware-accelerated-neuroevolution/index.html) Evolutionary computation has been shown to be a highly effective method. ### [Collective Intelligence for Deep Learning: A Survey of Recent Developments](/content/publication/collective-intelligence-for-deep-learning-a-survey-of-recent-developments/index.html) In the past decade, we have witnessed the rise of deep learning to dominate. ### [Sketch-based Creativity Support Tools using Deep Learning](/content/publication/sketch-based-creativity-support-tools-using-deep-learning/index.html) Sketching is a natural and effective visual communication medium. ### [Modern Evolution Strategies for Creativity: Fitting Concrete Images and Abstract Concepts](/content/publication/modern-evolution-strategies-for-creativity-fitting-concrete-images-and-abstract-concepts/index.html) Evolutionary algorithms have been used in the digital art scene since. ### [The Sensory Neuron as a Transformer: Permutation-Invariant Neural Networks for Reinforcement Learning](/content/publication/the-sensory-neuron-as-a-transformer-permutation-invariant-neural-networks-for-reinforcement-learning/index.html) In complex systems, we often observe complex global behavior. ### [Finding Game Levels with the Right Difficulty in a Few Trials through Intelligent Trial-and-Error](/content/publication/finding-game-levels-with-the-right-difficulty-in-a-few-trials-through-intelligent-trial-and-error/index.html) Methods for dynamic difficulty adjustment allow games to be tailored. ### [Scones: Towards Conversational Authoring of Sketches](/content/publication/scones-towards-conversational-authoring-of-sketches/index.html) Iteratively refining and critiquing sketches are crucial steps. ### [Neuroevolution of Self-Interpretable Agents](/content/publication/neuroevolution-of-self-interpretable-agents/index.html) Inattentional blindness is a psychological phenomenon that causes. ### [SketchTransfer: A Challenging New Task for Exploring Detail-Invariance and the Abstractions Learned by Deep Networks](/content/publication/sketchtransfer-a-challenging-new-task-for-exploring-detail-invariance-and-the-abstractions-learned-by-deep-networks/index.html) Deep networks have achieved excellent results in perceptual tasks. ### [Learning to Predict Without Looking Ahead: World Models Without Forward Prediction](/content/publication/learning-to-predict-without-looking-ahead-world-models-without-forward-prediction/index.html) Much of model-based reinforcement learning involves learning a model. ### [Weight Agnostic Neural Networks](/content/publication/weight-agnostic-neural-networks/index.html) Not all neural network architectures are created equal. ### [A Learned Representation for Scalable Vector Graphics](/content/publication/a-learned-representation-for-scalable-vector-graphics/index.html) Dramatic advances in generative models have resulted in near photographic. ### [Deep Learning for Classical Japanese Literature](/content/publication/deep-learning-for-classical-japanese-literature/index.html) Much of machine learning research focuses on producing models. ### [Learning Latent Dynamics for Planning from Pixels](/content/publication/learning-latent-dynamics-for-planning-from-pixels/index.html) Planning has been very successful for control tasks. ### [Reinforcement Learning for Improving Agent Design](/content/publication/reinforcement-learning-for-improving-agent-design/index.html) In many reinforcement learning tasks, the goal is to learn a policy. ### [Pommerman: A Multi-Agent Playground](/content/publication/pommerman-a-multi-agent-playground/index.html) We present Pommerman, a multi-agent environment based on the classic. ### [Recurrent World Models Facilitate Policy Evolution](/content/publication/recurrent-world-models-facilitate-policy-evolution/index.html) A generative recurrent neural network is quickly trained in. ### [World Models](/content/publication/world-models/index.html) We explore building generative neural network models of popular. ### [Learning via social awareness: improving sketch representations with facial feedback](/content/publication/learning-via-social-awareness-improving-sketch-representations-with-facial-feedback/index.html) In the quest towards general artificial intelligence (AI). ### [A Neural Representation of Sketch Drawings](/content/publication/a-neural-representation-of-sketch-drawings/index.html) We present sketch-rnn, a recurrent neural network (RNN) able to construct. ### [PathNet: Evolution Channels Gradient Descent in Super Neural Networks](/content/publication/pathnet-evolution-channels-gradient-descent-in-super-neural-networks/index.html) For artificial general intelligence (AGI) it would be efficient.