The design of the nanostructures that are used in the field of nano-photonics has remained complex, very often relying on the intuition and expertise of the designer, ultimately limiting the reach and penetration of this groundbreaking approach. Recently, there has been an increasing number of studies suggesting to apply Machine Learning techniques for the design of nanostructures. Most of these studies engage Deep Learning techniques, which entails training a Deep Neural Network (DNN) to approximate the highly non-linear function of the underlying physical process between spectra and nanostructures. At the end of the training, the DNN allows an on-demand design of nanostructures, i.e. the model can infer nanostructure geometries for desired spectra. In this work, we introduce spectra2pix, which is a model DNN trained to generate 2D images of the designed nanostructures. Our model architecture is not limited to a closed set of nanostructure shapes, and can be trained for the design of any geometry. We show, for the first time, a successful generalization ability by designing a completely unseen sub-family of geometries. This generalization capability highlights the importance of our model architecture, and allows higher applicability for real-world design problems.

Pre-trained word embeddings encode general word semantics and lexical regularities of natural language, and have proven useful across many NLP tasks, including word sense disambiguation, machine translation, and sentiment analysis, to name a few. In supervised tasks such as multiclass text classification (the focus of this article) it seems appealing to enhance word representations with ad-hoc embeddings that encode task-specific information. We propose (supervised) word-class embeddings (WCEs), and show that, when concatenated to (unsupervised) pre-trained word embeddings, they substantially facilitate the training of deep-learning models in multiclass classification by topic. We show empirical evidence that WCEs yield a consistent improvement in multiclass classification accuracy, using four popular neural architectures and six widely used and publicly available datasets for multiclass text classification. Our code that implements WCEs is publicly available at https://github.com/AlexMoreo/word-class-embeddings

Noname manuscript No. (will be inserted by the editor) Abstract We formalize networks with evolving structures as temporal networks and propose a generative link prediction model, Generative Link Sequence Modeling (GLSM), to predict future links for temporal networks. GLSM captures the temporal link formation patterns from the observed links with a sequence modeling framework and has the ability to generate the emerging links by inferring from the probability distribution on the potential future links. To avoid overfitting caused by treating each link as a unique token, we propose a self-tokenization mechanism to transform each raw link in the network to an abstract aggregation token automatically. The self-tokenization is seamlessly integrated into the sequence modeling framework, which allows the proposed GLSM model to have the generalization capability to discover link formation patterns beyond raw link sequences. We compare GLSM with the existing state-of-art methods on five real-world datasets. The experimental results demonstrate that GLSM obtains future positive links effectively in a generative fashion while achieving the best performance (2-10% improvements on AUC) among other alternatives. Keywords Temporal link prediction, sequence modeling, recurrent neural network, self-tokenization mechanism 1 Introduction Many real-world applications could be modeled as link prediction problems. Lu Bai is the corresponding author, Email: bailucs@cufe.edu.cn 1. Central University of Finance and Economics, Beijing, P.R. China. Two mainstream categories in link prediction are either based on the statistical patterns of the link formation behaviors of the network [10, 2, 17] or the graph representation learning [31, 33] methods which embed nodes as vectors with respect to the network topological information. Most of these methods are discriminative models that verify whether an unknown link given during the test time is rational by training a classifier on existing links and negative samples [19].

For an unknown (new) classification dataset, choosing an appropriate deep learning architecture is often a recursive, time-taking, and laborious process. In this research, we propose a novel technique to recommend a suitable architecture from a repository of known models. Further, we predict the performance accuracy of the recommended architecture on the given unknown dataset, without the need for training the model. We propose a model encoder approach to learn a fixed length representation of deep learning architectures along with its hyperparameters, in an unsupervised fashion. We manually curate a repository of image datasets with corresponding known deep learning models and show that the predicted accuracy is a good estimator of the actual accuracy. We discuss the implications of the proposed approach for three benchmark images datasets and also the challenges in using the approach for text modality. To further increase the reproducibility of the proposed approach, the entire implementation is made publicly available along with the trained models.

--Powerful generative adversarial networks (GAN) have been developed to automatically synthesize realistic images from text. However, most existing tasks are limited to generating simple images such as flowers from captions. In this work, we extend this problem to the less addressed domain of face generation from fine-grained textual descriptions of face, e.g., "A person has curly hair, oval face, and mustache" . We are motivated by the potential of automated face generation to impact and assist critical tasks such as criminal face reconstruction. Since current datasets for the task are either very small or do not contain captions, we generate captions for images in the CelebA dataset by creating an algorithm to automatically convert a list of attributes to a set of captions. We then model the highly multi-modal problem of text to face generation as learning the conditional distribution of faces (conditioned on text) in same latent space. We utilize the current state-of-the-art GAN (DC-GAN with GAN-CLS loss) for learning conditional multi-modality. The presence of more fine-grained details and variable length of the captions makes the problem easier for a user but more difficult to handle compared to the other text-to-image tasks. We flipped the labels for real and fake images and added noise in discriminator . Generated images for diverse textual descriptions show promising results. In the end, we show how the widely used inceptions score is not a good metric to evaluate the performance of generative models used for synthesizing faces from text. I NTRODUCTION Photographic text-to-face synthesis is a mainstream problem with potential applications in image editing, video games, or for accessibility.

Learning representations of data is an important problem in statistics and machine learning. While the origin of learning representations can be traced back to factor analysis and multidimensional scaling in statistics, it has become a central theme in deep learning with important applications in computer vision and computational neuroscience. In this article, we review recent advances in learning representations from a statistical perspective. In particular, we review the following two themes: (a) unsupervised learning of vector representations and (b) learning of both vector and matrix representations.

Recent advances in adversarial attacks uncover the intrinsic vulnerability of modern deep neural networks. Since then, extensive efforts have been devoted to enhancing the robustness of deep networks via specialized learning algorithms and loss functions. In this work, we take an architectural perspective and investigate the patterns of network architectures that are resilient to adversarial attacks. To obtain the large number of networks needed for this study, we adopt one-shot neural architecture search, training a large network for once and then finetuning the sub-networks sampled therefrom. The sampled architectures together with the accuracies they achieve provide a rich basis for our study. Our "robust architecture Odyssey" reveals several valuable observations: 1) densely connected patterns result in improved robustness; 2) under computational budget, adding convolution operations to direct connection edge is effective; 3) flow of solution procedure (FSP) matrix is a good indicator of network robustness. Based on these observations, we discover a family of robust architectures (RobNets). On various datasets, including CIFAR, SVHN, and Tiny-ImageNet, RobNets exhibit superior robustness performance to other widely used architectures. Notably, RobNets substantially improve the robust accuracy (~5% absolute gains) under both white-box and black-box attacks, even with fewer parameter numbers.

Transfer learning is becoming the de facto solution for vision and text encoders in the front-end processing of machine learning solutions. Utilizing vast amounts of knowledge in pre-trained models and subsequent fine-tuning allows achieving better performance in domains where labeled data is limited. In this paper, we analyze the efficiency of transfer learning in visual reasoning by introducing a new model (SAMNet) and testing it on two datasets: COG and CLEVR. Our new model achieves state-of-the-art accuracy on COG and shows significantly better generalization capabilities compared to the baseline. We also formalize a taxonomy of transfer learning for visual reasoning around three axes: feature, temporal, and reasoning transfer. Based on extensive experimentation of transfer learning on each of the two datasets, we show the performance of the new model along each axis.

Fictitious play with reinforcement learning is a general and effective framework for zero-sum games. However, using the current deep neural network models, the implementation of fictitious play faces crucial challenges. Neural network model training employs gradient descent approaches to update all connection weights, and thus is easy to forget the old opponents after training to beat the new opponents. Existing approaches often maintain a pool of historical policy models to avoid the forgetting. However, learning to beat a pool in stochastic games, i.e., a wide distribution over policy models, is either sample-consuming or insufficient to exploit all models with limited amount of samples. In this paper, we propose a learning process with neural fictitious play to alleviate the above issues. We train a single model as our policy model, which consists of sub-models and a selector. Everytime facing a new opponent, the model is expanded by adding a new sub-model, where only the new sub-model is updated instead of the whole model. At the same time, the selector is also updated to mix up the new sub-model with the previous ones at the state-level, so that the model is maintained as a behavior strategy instead of a wide distribution over policy models. Experiments on Kuhn poker, a grid-world Treasure Hunting game, and Mini-RTS environments show that the proposed approach alleviates the forgetting problem, and consequently improves the learning efficiency and the robustness of neural fictitious play.

The unification of low-level perception and high-level reasoning is a long-standing problem in artificial intelligence, which has the potential to not only bring the areas of logic and learning closer together but also demonstrate how abstract concepts might emerge from sensory data. Precisely because deep learning methods dominate perception-based learning, including vision, speech, and linguistic grammar, there is fast-growing literature on how to integrate symbolic reasoning and deep learning. Broadly, efforts seem to fall into three camps: those focused on defining a logic whose formulas capture deep learning, ones that integrate symbolic constraints in deep learning, and others that allow neural computations and symbolic reasoning to co-exist separately, to enjoy the strengths of both worlds. In this paper, we identify another dimension to this inquiry: what do the hidden layers really capture, and how can we reason about that logically? In particular, we consider autoencoders that are widely used for dimensionality reduction and inject a symbolic generative framework onto the feature layer. This allows us, among other things, to generate example images for a class to get a sense of what was learned. Moreover, the modular structure of the proposed model makes it possible to learn relations over multiple images at a time, as well as handle noisy labels. Our empirical evaluations show the promise of this inquiry.