Coline Devin

Learn More
We address the problem of adapting robotic perception from simulated to real-world environments. For many robotic control tasks, real training imagery is expensive to obtain, but a large amount of synthetic data is easy to generate through simulation. We propose a method that adapts representations using a small number of paired synthetic and real views of(More)
Neural language models learn word representations, or embeddings, that capture rich linguistic and conceptual information. Here we investigate the embeddings learned by neural machine translation models, a recently-developed class of neu-ral language model. We show that embeddings from translation models outper-form those learned by monolingual models at(More)
Neural language models learn word representations that capture rich linguistic and conceptual information. Here we investigate the embeddings learned by neural machine translation models. We show that translation-based embeddings outper-form those learned by cutting-edge monolingual models at single-language tasks requiring knowledge of conceptual(More)
— Reinforcement learning (RL) can automate a wide variety of robotic skills, but learning each new skill requires considerable real-world data collection and manual representation engineering to design policy classes or features. Using deep reinforcement learning to train general purpose neural network policies alleviates some of the burden of manual(More)
Analogs of bombesin in which the peptide bond between the two last amino acid residues were replaced by a pseudopeptide bond mimicking the transition state analog were evaluated. These compounds were able to recognize the bombesin receptor on isolated rat pancreatic acini with high potency (Ki from 0.60 +/- 0.27 nM to 4.3 +/- 2.3 nM). Although they were(More)
People can learn a wide range of tasks from their own experience, but can also learn from observing other creatures. This can accelerate acquisition of new skills even when the observed agent differs substantially from the learning agent in terms of morphology. In this paper, we examine how reinforcement learning algorithms can transfer knowledge between(More)
The peptides of the bombesin family are involved in stimulation of mitogenesis in various cell lines, including cancerous cell lines. Bombesin receptor antagonists are of great interest to inhibit this proliferation. We have synthesized a potent bombesin receptor antagonist, e.g., compound JMV641 [H-DPhe-Gln-Trp-Ala-Val-Gly-His-NH-*CH[CH2-CH(CH3)2]-**CHOH-(More)
  • 1