In this paper, we propose a method for image-set classification based on convex cone models, focusing on the effectiveness of convolutional neural network (CNN) features as inputs. CNN features have non-negative values when using the rectified linear unit as an activation function. This naturally leads us to model a set of CNN features by a convex cone and measure the geometric similarity of convex cones for classification. To establish this framework, we sequentially define multiple angles between two convex cones by repeating the alternating least squares method and then define the geometric similarity between the cones using the obtained angles. Moreover, to enhance our method, we introduce a discriminant space, maximizing the between-class variance (gaps) and minimizes the within-class variance of the projected convex cones onto the discriminant space, similar to a Fisher discriminant analysis. Finally, classification is based on the similarity between projected convex cones. The effectiveness of the proposed method was demonstrated experimentally using a private, multi-view hand shape dataset and two public databases.

Most artificial intelligence models have limiting ability to solve new tasks faster, without forgetting previously acquired knowledge. The recently emerging paradigm of continual learning aims to solve this issue, in which the model learns various tasks in a sequential fashion. In this work, a novel approach for continual learning is proposed, which searches for the best neural architecture for each coming task via sophisticatedly designed reinforcement learning strategies. We name it as Reinforced Continual Learning. Our method not only has good performance on preventing catastrophic forgetting but also fits new tasks well. The experiments on sequential classification tasks for variants of MNIST and CIFAR-100 datasets demonstrate that the proposed approach outperforms existing continual learning alternatives for deep networks.

The success of deep learning has been due in no small part to the availability of large annotated datasets. Thus, a major bottleneck in the current learning pipeline is the human annotation of data, which can be quite time consuming. For a given problem setting, we aim to circumvent this issue via the use of an externally specified energy function appropriate for that setting; we call this the "Deep Energy" approach. We show how to train a network on an entirely unlabelled dataset using such an energy function, and apply this general technique to learn CNNs for two specific tasks: seeded segmentation and image matting. Once the network parameters have been learned, we obtain a high-quality solution in a fast feed-forward style, without the need to repeatedly optimize the energy function for each image.

Many mobile robots rely on 2D laser scanners for localization, mapping, and navigation. However, those sensors are unable to correctly provide distance to obstacles such as glass panels and tables whose actual occupancy is invisible at the height the sensor is measuring. In this work, instead of estimating the distance to obstacles from richer sensor readings such as 3D lasers or RGBD sensors, we present a method to estimate the distance directly from raw 2D laser data. To learn a mapping from raw 2D laser distances to obstacle distances we frame the problem as a learning task and train a neural network formed as an autoencoder. A novel configuration of network hyperparameters is proposed for the task at hand and is quantitatively validated on a test set. Finally, we qualitatively demonstrate in real time on a Care-O-bot 4 that the trained network can successfully infer obstacle distances from partial 2D laser readings.

Robustness to adversarial perturbation has become an extremely important criterion for applications of machine learning in security-sensitive domains such as spam detection [25], fraud detection [6], criminal justice [3], malware detection [13], and financial markets [27]. Systematic methods for generating adversarial examples by small perturbations of original input data, also known as "attack," have been developed to operationalize this criterion and to drive the development of more robust learning systems [4, 26, 7]. Most of the work in this area has focused on differentiable models with continuous input spaces [26, 7, 14, 14]. In this setting, the proposed attack strategies add a gradient-based perturbation to the original input. It has been shown that such perturbations can result in a dramatic decrease in the predictive accuracy of the model. Thus this line of research has demonstrated the vulnerability of deep neural networks to adversarial examples in tasks like image classification and speech recognition. We focus instead on adversarial attacks on models with discrete input data, such as text data, where each feature of an input sample has a categorical domain. While gradient-based approaches are not directly applicable to this setting, variations of gradient-based approaches have been shown effective in differentiable models. For example, Li et al. [15] proposed to locate the top features with the largest gradient magnitude of their embedding, and Papernot et al. [20] proposed to modify randomly selected features of an input through perturbing each feature by signs of the gradient, and project them onto the closest vector in the embedding space.

Self-play (Sukhbaatar et al., 2017) is an unsupervised training procedure which enables the reinforcement learning agents to explore the environment without requiring any external rewards. We augment the self-play setting by providing an external memory where the agent can store experience from the previous tasks. This enables the agent to come up with more diverse self-play tasks resulting in faster exploration of the environment. The agent pretrained in the memory augmented self-play setting easily outperforms the agent pretrained in no-memory self-play setting.

Kernel Methods Next Generation (KMNG) introduces a framework to use kernel approximates in the mini-batch setting with SGD Optimizer as an alternative to Deep Learning. McKernel is a C++ library for KMNG ML Large-scale. It contains a CPU optimized implementation of the Fastfood algorithm that allows the computation of approximated kernel expansions in log-linear time. The algorithm requires to compute the product of Walsh Hadamard Transform (WHT) matrices. A cache friendly SIMD Fast Walsh Hadamard Transform (FWHT) that achieves compelling speed and outperforms current state-of-the-art methods has been developed. McKernel allows to obtain non-linear classification combining Fastfood and a linear classifier.

Convolutional neural networks are deep artificial neural networks that are used primarily to classify images (e.g. They are algorithms that can identify faces, individuals, street signs, tumors, platypuses and many other aspects of visual data. Convolutional networks perform optical character recognition (OCR) to digitize text and make natural-language processing possible on analog and hand-written documents, where the images are symbols to be transcribed. CNNs can also be applied to sound when it is represented visually as a spectrogram. More recently, convolutional networks have been applied directly to text analytics as well as graph data with graph convolutional networks.

Deep learning: By learning from images, audio, and text, data deep neural networks can mimic the human brain. They have been in use for more than a decade, but there's a lot still to be discovered and study how a neural network learns and how can they be made effective. Deep learning is getting smarter though, instead of feeding hundreds and thousands of data points, today's systems can give an accurate output by factoring only few hundred data points.Deep reinforcement learning: The developers of artificial intelligence are applying the principle of reinforced learning to their systems. This is the reason why the well-known AlphaGo was able to beat a human champion. Lately, it beat a DOTA champion in a very composite game by educating itself to play the game within two weeks.

For the first time, new research suggests artificial intelligence may be better than highly-trained humans at detecting skin cancer. A study conducted by an international team of researchers pitted experienced dermatologists against a machine learning system, known as a deep learning convolutional neural network, or CNN, to see which was more effective at detecting malignant melanomas. The results? "Most dermatologists were outperformed by the CNN," the researchers wrote in their report, published in the journal Annals of Oncology. Fifty-eight dermatologists from 17 countries around the world participated in the study. More than half of the doctors were considered expert level with more than five years' experience.