A biologically plausible neural network model named Perception Coordination Network (PCN) is proposed for online multi-modal concept acquisition and binding. It is a hierarchical structure inspired by the structure of the brain, and functionally divided into the primary sensory area (PSA), the primary sensory association area (SAA), and the higher order association area (HAA). The PSA processes many elementary features, e.g., colors, shapes, syllables, and basic flavors, etc. The SAA combines these elementary features to represent the unimodal concept of an object, e.g., the image, name and taste of an apple, etc. The HAA connects several primary sensory association areas like a function of synaesthesia, which means associating the image, name and taste of an object. PCN is able to continuously acquire and bind multi-modal concepts in an online way. Experimental results suggest that PCN can handle the multi-modal concept acquisition and binding problem effectively.

This paper proposes a cooperative learning algorithm to train both the undirected energy-based model and the directed latent variable model jointly. The learning algorithm interweaves the maximum likelihood algorithms for learning the two models, and each iteration consists of the following two steps: (1) Modified contrastive divergence for energy-based model: The learning of the energy-based model is based on the contrastive divergence, but the finite-step MCMC sampling of the model is initialized from the synthesized examples generated by the latent variable model instead of being initialized from the observed examples. (2) MCMC teaching of the latent variable model: The learning of the latent variable model is based on how the MCMC in (1) changes the initial synthesized examples generated by the latent variable model, where the latent variables that generate the initial synthesized examples are known so that the learning is essentially supervised. Our experiments show that the cooperative learning algorithm can learn realistic models of images.

In this paper, we are interested in designing small CNNs by decoupling the convolution along the spatial and channel domains. Most existing decoupling techniques focus on approximating the filter matrix through decomposition. In contrast, we provide a two-step interpretation of the standard convolution from the filter at a single location to all locations, which is exactly equivalent to the standard convolution. Motivated by the observations in our decoupling view, we propose an effective approach to relax the sparsity of the filter in spatial aggregation by learning a spatial configuration, and reduce the redundancy by reducing the number of intermediate channels. Our approach achieves comparable classification performance with the standard uncoupled convolution, but with a smaller model size over CIFAR-100, CIFAR-10 and ImageNet.

Effective methods for learning deep neural networks with fewer parameters are urgently required, since storage and computations of heavy neural networks have largely prevented their widespread use on mobile devices. Compared with algorithms which directly remove weights or filters for obtaining considerable compression and speed-up ratios, training thin deep networks exploiting the student-teacher learning paradigm is more flexible. However, it is very hard to determine which formulation is optimal to measure the information inherited from teacher networks. To overcome this challenge, we utilize the generative adversarial network (GAN) to learn the student network. In practice, the generator is exactly the student network with extremely less parameters and the discriminator is used as a teaching assistant for distinguishing features extracted from student and teacher networks. By simultaneously optimizing the generator and the discriminator, the resulting student network can produce features of input data with the similar distribution as that of features of the teacher network. Extensive experimental results on benchmark datasets demonstrate that the proposed method is capable of learning well-performed portable networks, which is superior to the state-of-the-art methods.

Hardware accelerations of deep learning systems have been extensively investigated in industry and academia. The aim of this paper is to achieve ultra-high energy efficiency and performance for hardware implementations of deep neural networks (DNNs). An algorithm-hardware co-optimization framework is developed, which is applicable to different DNN types, sizes, and application scenarios. The algorithm part adopts the general block-circulant matrices to achieve a fine-grained tradeoff of accuracy and compression ratio. It applies to both fully-connected and convolutional layers and contains a mathematically rigorous proof of the effectiveness of the method. The proposed algorithm reduces computational complexity per layer from O(n 2 ) to O(n log n) and storage complexity from O(n 2 ) to O(n), both for training and inference. The hardware part consists of highly efficient Field Programmable Gate Array (FPGA)-based implementations using effective reconfiguration, batch processing, deep pipelining, resource re-using, and hierarchical control. Experimental results demonstrate that the proposed framework achieves at least 152X speedup and 71X energy efficiency gain compared with IBM TrueNorth processor under the same test accuracy. It achieves at least 31X energy efficiency gain compared with the reference FPGA-based work.

Predictions of time-series are widely used in different disciplines. We propose CoR, Sparse Gaussian Conditional Random Fields (SGCRF) on top of Recurrent Neural Networks (RNN), for problems of this kind. CoR gains advantages from both RNN and SGCRF. It can not only effectively represent the temporal correlations in observed data, but can also learn the structured information of the output. CoR is challenging to train because it is a hybrid of deep neural networks and densely-connected graphical models. Alternative training can be a tractable way to train CoR, and furthermore, an end-to-end training method is proposed to train CoR more efficiently. CoR is evaluated by both synthetic data and real-world data, and it shows a significant improvement in performance over state-of-the-art methods.

We present a deep generative model for Zero-Shot Learning (ZSL). Unlike most existing methods for this problem, that represent each class as a point (via a semantic embedding), we represent each seen/unseen class using a class-specific latent-space distribution, conditioned on class attributes. We use these latent-space distributions as a prior for a supervised variational autoencoder (VAE), which also facilitates learning highly discriminative feature representations for the inputs. The entire framework is learned end-to-end using only the seen-class training data. At test time, the label for an unseen-class test input is the class that maximizes the VAE lower bound. We further extend the model to a (i) semi-supervised/transductive setting by leveraging unlabeled unseen-class data via an unsupervised learning module, and (ii) few-shot learning where we also have a small number of labeled inputs from the unseen classes. We compare our model with several state-of-the-art methods through a comprehensive set of experiments on a variety of benchmark data sets.

Sum-Product Networks (SPNs) are a deep probabilistic architecture that up to now has been successfully employed for tractable inference. Here, we extend their scope towards unsupervised representation learning: we encode samples into continuous and categorical embeddings and show that they can also be decoded back into the original input space by leveraging MPE inference. We characterize when this Sum-Product Autoencoding (SPAE) leads to equivalent reconstructions and extend it towards dealing with missing embedding information. Our experimental results on several multi-label classification problems demonstrate that SPAE is competitive with state-of-the-art autoencoder architectures, even if the SPNs were never trained to reconstruct their inputs.

Gaussian Processes (GPs) provide an extremely powerful mechanism to model a variety of problems but incur an O(N 3 ) complexity in the number of data samples. Common approximation methods rely on what are often termed inducing points but still typically incur an O(NM 2 ) complexity in the data and corresponding inducing points. Using Random Fourier Feature (RFF) maps, we overcome this by transforming the problem into a Bayesian Linear Regression formulation upon which we apply a Bayesian Variational treatment that also allows learning the corresponding kernel hyperparameters, likelihood and noise parameters. In this paper we introduce an alternative method using Fourier series to obtain spectral representations of common kernels, in particular for periodic warpings, which surprisingly have a convergent, non-random form using special functions, requiring fewer spectral features to approximate their corresponding kernel to high accuracy. Using this, we can fuse the Random Fourier Feature spectral representations of common kernels with their periodic counterparts to show how they can more effectively and expressively learn patterns in time-series for both interpolation and extrapolation. This method combines robustness, scalability and equally importantly, interpretability through a symbolic declarative grammar that is both functionally and humanly intuitive — a property that is crucial for explainable decision making. Using probabilistic programming and Variational Inference we are able to efficiently optimise over these rich functional representations. We show significantly improved Gram matrix approximation errors, and also demonstrate the method in several time-series problems comparing other commonly used approaches such as recurrent neural networks.

Deep Neural Networks (DNNs) have demonstrated remarkable performance in a diverse range of applications. Along with the prevalence of deep learning, it has been revealed that DNNs are vulnerable to attacks. By deliberately crafting adversarial examples, an adversary can manipulate a DNN to generate incorrect outputs, which may lead catastrophic consequences in applications such as disease diagnosis and self-driving cars. In this paper, we propose an effective method to detect adversarial examples in image classification. Our key insight is that adversarial examples are usually sensitive to certain image transformation operations such as rotation and shifting. In contrast, a normal image is generally immune to such operations. We implement this idea of image transformation and evaluate its performance in oblivious attacks. Our experiments with two datasets show that our technique can detect nearly 99% of adversarial examples generated by the state-of-the-art algorithm. In addition to oblivious attacks, we consider the case of white-box attacks. We propose to introduce randomness in the process of image transformation, which can achieve a detection ratio of around 70%.