人工智能资料库：第31辑（20170212）

作者：chen_h
微信号 & QQ：862251340
微信公众号：coderpai

---

1. 【博客】Neural network learns to select potential anticancer drugs

简介：

Scientists from Mail.Ru Group, Insilico Medicine and MIPT have for the first time applied a generative neural network to create new pharmaceutical medicines with the desired characteristics. By using Generative Adversarial Networks (GANs) developed and trained to “invent” new molecular structures, there may soon be a dramatic reduction in the time and cost of searching for substances with potential medicinal properties. The researchers intend to use these technologies in the search for new medications within various areas from oncology to CVDs and even anti-infectives. The first results were submitted toОncotargetin June 2016 and spent several months in review. Since that time, the group has made many improvements to the system and engaged with some of the leading pharmaceutical companies.

原文链接：https://mipt.ru/english/news/neural_network_learns_to_select_potential_anticancer_drugs

---

2.【论文&视频】Active One-shot Learning

简介：

Recent advances in one-shot learning have produced models that can learn from a handful of labeled examples, for passive classification and regression tasks. This paper combines reinforcement learning with one-shot learning, allowing the model to decide, during classification, which examples are worth labeling. We introduce a classification task in which a stream of images are presented and, on each time step, a decision must be made to either predict a label or pay to receive the correct label. We present a recurrent neural network based action-value function, and demonstrate its ability to learn how and when to request labels. Through the choice of reward function, the model can achieve a higher prediction accuracy than a similar model on a purely supervised task, or trade prediction accuracy for fewer label requests.

原文链接：https://cs.stanford.edu/~woodward/papers/active_one_shot_learning_2016.pdf

视频链接：https://www.youtube.com/watch?v=CzQSQ_0Z-QU&feature=youtu.be

---

3.【博客】Reinforcement Learning for Photonic Engineering

简介：

The paper I will comment and review today here for this blog is quite special. It is about a topic I cherish – Photonics and Optical Engineering once was my career path… -, but it is double dose of reinforced engagement. It meshes Photonic Engineering with the techniques and conceptual framework of Reinforcement Learning. Reinforcement Learning is one of the most important and significant fields of development within the broader Machine Learning and Computer Sciencesubjects. Applied, as it is wonderfully described in this paper (that tastes more like a scientific and technological essay, and how close to my inclinations that could be…) to the topic of light transport simulation, it showed capacity to improve this photonic engineering challenge, possibly providing a path to further enhancement to rendering of images inComputer Graphics.

原文链接：https://theinformationageblog.wordpress.com/2017/02/10/reinforcement-learning-for-photonic-engineering/

---

4.【代码】Gaussian processes framework in python

简介：

A Gaussian process framework in python

原文链接：https://github.com/sheffieldML/GPy

---

5.【博客】Hyperparameter optimization with approximate gradient

简介：

Most machine learning models rely on at least one hyperparameter to control for model complexity. For example, logistic regression commonly relies on a regularization parameter that controls the amount of ℓ2ℓ2 regularization. Similarly, kernel methods also have hyperparameters that control for properties of the kernel, such as the “width” parameter in the RBF kernel. The fundamental distinction between model parameters and hyperparameters is that, while model parameters are estimated by minimizing a goodness of fit with the training data, hyperparameters need to be estimated by other means (such as a cross-validation loss), as otherwise models with excessive would be selected, a phenomenon known as overfitting.

原文链接：http://fa.bianp.net/blog/2016/hyperparameter-optimization-with-approximate-gradient/

---