We describe an end-to-end neural network weight compression approach that draws inspiration from recent latent-variable data compression methods. The network parameters (weights and biases) are represented in a "latent" space, amounting to a reparameterization. This space is equipped with a learned probability model, which is used to impose an entropy penalty on the parameter representation during training, and to compress the representation using arithmetic coding after training. We are thus maximizing accuracy and model compressibility jointly, in an end-to-end fashion, with the rate--error trade-off specified by a hyperparameter. We evaluate our method by compressing six distinct model architectures on the MNIST, CIFAR-10 and ImageNet classification benchmarks. Our method achieves state-of-the-art compression on VGG-16, LeNet300-100 and several ResNet architectures, and is competitive on LeNet-5.

We introduce a multimodal visual-textual search refinement method for fashion garments. Existing search engines do not enable intuitive, interactive, refinement of retrieved results based on the properties of a particular product. We propose a method to retrieve similar items, based on a query item image and textual refinement properties. We believe this method can be leveraged to solve many real-life customer scenarios, in which a similar item in a different color, pattern, length or style is desired. We employ a joint embedding training scheme in which product images and their catalog textual metadata are mapped closely in a shared space. This joint visual-textual embedding space enables manipulating catalog images semantically, based on textual refinement requirements. We propose a new training objective function, Mini-Batch Match Retrieval, and demonstrate its superiority over the commonly used triplet loss. Additionally, we demonstrate the feasibility of adding an attribute extraction module, trained on the same catalog data, and demonstrate how to integrate it within the multimodal search to boost its performance. We introduce an evaluation protocol with an associated benchmark, and compare several approaches.

Recent advances in multi-modal vision and language tasks enable a new set of applications. In this paper, we consider the task of generating natural language fashion feedback on outfit images. We collect a unique dataset, which contains outfit images and corresponding positive and constructive fashion feedback. We treat each feedback type separately, and train deep generative encoder-decoder models with visual attention, similar to the standard image captioning pipeline. Following this approach, the generated sentences tend to be too general and noninformative. We propose an alternative decoding technique based on the Maximum Mutual Information objective function, which leads to more diverse and detailed responses. We evaluate our model with common language metrics, and also show human evaluation results. This technology is applied within the "Alexa, how do I look?" feature, publicly available in Echo Look devices.

Technological breakthroughs on smart homes, self-driving cars, health care and robotic assistants, in addition to reinforced law regulations, have critically influenced academic research on explainable machine learning. A sufficient number of researchers have implemented ways to explain indifferently any black box model for classification tasks. A drawback of building agnostic explanators is that the neighbourhood generation process is universal and consequently does not guarantee true adjacency between the generated neighbours and the instance. This paper explores a methodology on providing explanations for a neural network's decisions, in a local scope, through a process that actively takes into consideration the neural network's architecture on creating an instance's neighbourhood, that assures the adjacency among the generated neighbours and the instance.

Graph clustering is a fundamental task which discovers communities or groups in networks. Recent studies have mostly focused on developing deep learning approaches to learn a compact graph embedding, upon which classic clustering methods like k-means or spectral clustering algorithms are applied. These two-step frameworks are difficult to manipulate and usually lead to suboptimal performance, mainly because the graph embedding is not goal-directed, i.e., designed for the specific clustering task. In this paper, we propose a goal-directed deep learning approach, Deep Attentional Embedded Graph Clustering (DAEGC for short). Our method focuses on attributed graphs to sufficiently explore the two sides of information in graphs. By employing an attention network to capture the importance of the neighboring nodes to a target node, our DAEGC algorithm encodes the topological structure and node content in a graph to a compact representation, on which an inner product decoder is trained to reconstruct the graph structure. Furthermore, soft labels from the graph embedding itself are generated to supervise a self-training graph clustering process, which iteratively refines the clustering results. The self-training process is jointly learned and optimized with the graph embedding in a unified framework, to mutually benefit both components. Experimental results compared with state-of-the-art algorithms demonstrate the superiority of our method.

Global optimization problems whose objective function is expensive to evaluate can be solved effectively by recursively fitting a surrogate function to function samples and minimizing an acquisition function to generate new samples. The acquisition step trades off between seeking for a new optimization vector where the surrogate is minimum (exploitation of the surrogate) and looking for regions of the feasible space that have not yet been visited and that may potentially contain better values of the objective function (exploration of the feasible space). This paper proposes a new global optimization algorithm that uses a combination of inverse distance weighting (IDW) and radial basis functions (RBF) to construct the acquisition function. Rather arbitrary constraints that are simple to evaluate can be easily taken into account by the approach. Compared to Bayesian optimization, the proposed algorithm is computationally lighter and, as we show in a set of benchmark global optimization and hyperparameter tuning problems, it has a very similar (and sometimes superior) performance. MATLAB and Python implementations of the proposed approach are available at http://cse.lab.imtlucca.it/~bemporad/idwgopt

Deep quantization of neural networks (below eight bits) offers significant promise in reducing their compute and storage cost. Albeit alluring, without special techniques for training and optimization, deep quantization results in significant accuracy loss. To further mitigate this loss, we propose a novel sinusoidal regularization, called SinReQ, for deep quantized training. SinReQ adds a periodic term to the original objective function of the underlying training algorithm. SinReQ exploits the periodicity, differentiability, and the desired convexity profile in sinusoidal functions to automatically propel weights towards values that are inherently closer to quantization levels. Since, this technique does not require invasive changes to the training procedure, SinReQ can harmoniously enhance quantized training algorithms. SinReQ offers generality and flexibility as it is not limited to a certain bitwidth or a uniform assignment of bitwidths across layers. We carry out experimentation using the CIFAR-10, ResNet-20, SVHN DNNs with three to five bits for quantization and show the versatility of SinReQ in enhancing multiple quantized training algorithms, DoReFa (Zhou et al., 2016) and WRPN (Mishra et al., 2018). Averaging across all the bit configurations shows that SinReQ closes the accuracy gap between these two techniques and the full-precision runs by 35.7% and 37.1%, respectively. That is improving the absolute accuracy of DoReFa and WRPN up to 5.3% and 2.6%, respectively.

While model-based deep reinforcement learning (RL) holds great promise for sample efficiency and generalization, learning an accurate dynamics model is often challenging and requires substantial interaction with the environment. A wide variety of domains have dynamics that share common foundations like the laws of classical mechanics, which are rarely exploited by existing algorithms. In fact, humans continuously acquire and use such dynamics priors to easily adapt to operating in new environments. In this work, we propose an approach to learn task-agnostic dynamics priors from videos and incorporate them into an RL agent. Our method involves pre-training a frame predictor on task-agnostic physics videos to initialize dynamics models (and fine-tune them) for unseen target environments. Our frame prediction architecture, SpatialNet, is designed specifically to capture localized physical phenomena and interactions. Our approach allows for both faster policy learning and convergence to better policies, outperforming competitive approaches on several different environments. We also demonstrate that incorporating this prior allows for more effective transfer between environments.

Automated classifiers may be better than physicians when it comes to diagnosing pigmented skin lesions, but human supervision is still needed, researchers found. All machine-learning algorithms reached a mean of 2.01 more correct diagnoses than did all human readers (17.91 vs 19.92; P 0.0001), reported Harald Kittler, MD, of the Medical University of Vienna in Austria, and colleagues in The Lancet Oncology. When comparing the top three machine learning algorithms with 27 human experts with over a decade of experience, the algorithms still outperformed the experts (18.78 vs 25.43; P 0.0001), the investigators found. Notably, the difference between the top three algorithms and experts was significantly lower for images that were gathered from centers that did not contribute images for the training set when compared with other image sets, although there was human under-performance once again (11.4% vs 3.6%; P 0.0001), the researchers wrote. In this study, machine-learning classifiers performed better than experienced human readers in the diagnosis of pigmented skin lesions, suggesting that machine learning should have a more important role in clinical practice, the investigators said.

The first Edge AI incubator came to a close at Talent Garden Dublin with entrepreneurs demonstrating their computer vision technologies. Intel's first ever Edge AI incubator programme, first announced in October 2018, offered a unique opportunity for a diverse group of companies to integrate computer vision technology into their hardware solutions, building on the expertise in machine vision developed by the Intel Movidius team. After what was deemed a fantastic response to the initial call for participants, the selected companies set to work within the incubator at Talent Garden Dublin in January 2019. 'We are looking at how we can take the learnings of this first incubator programme and use it to inform further opportunities for technology innovation' – NOEL MURPHY The technologies these Edge AI companies have been developing for the past six months are based on Myriad 2 and Myriad X hardware and software from Intel, with the support of the local Intel team in Leixlip. "The incubator programme has been about taking technology being developed by Intel Movidius and using it to catalyse innovation, which we hope will support Ireland in becoming a powerhouse for Edge AI," said Noel Murphy, vice-president of Intel's AIPG group and director of engineering for Intel Movidius.