We establish an equivalence between information bottleneck (IB) learning and an unconventional quantization problem, "IB quantization". Under this equivalence, standard neural network models correspond to scalar IB quantizers. We prove a coding theorem for IB quantization, which implies that scalar IB quantizers are in general inferior to vector IB quantizers. This inspires us to develop a learning framework for neural networks, AgrLearn, that corresponds to vector IB quantizers. We experimentally verify that AgrLearn applied to some deep network models of current art improves upon them, while requiring less training data. With a heuristic smoothing, AgrLearn further improves its performance, resulting in new state of the art in image classification on Cifar10.

We consider the problem of discovering the simplest latent variable that can make two observed discrete variables conditionally independent. This problem has appeared in the literature as probabilistic latent semantic analysis (pLSA), and has connections to non-negative matrix factorization. When the simplicity of the variable is measured through its cardinality, we show that a solution to this latent variable discovery problem can be used to distinguish direct causal relations from spurious correlations among almost all joint distributions on simple causal graphs with two observed variables. Conjecturing a similar identifiability result holds with Shannon entropy, we study a loss function that trades-off between entropy of the latent variable and the conditional mutual information of the observed variables. We then propose a latent variable discovery algorithm -- LatentSearch -- and show that its stationary points are the stationary points of our loss function. We experimentally show that LatentSearch can indeed be used to distinguish direct causal relations from spurious correlations.

An understanding of the nature of objects could help robots to solve both high-level abstract tasks and improve performance at lower-level concrete tasks. Although deep learning has facilitated progress in image understanding, a robot's performance in problems like object recognition often depends on the angle from which the object is observed. Traditionally, robot sorting tasks rely on a fixed top-down view of an object. By changing its viewing angle, a robot can select a more semantically informative view leading to better performance for object recognition. In this paper, we introduce the problem of semantic view selection, which seeks to find good camera poses to gain semantic knowledge about an observed object. We propose a conceptual formulation of the problem, together with a solvable relaxation based on clustering. We then present a new image dataset consisting of around 10k images representing various views of 144 objects under different poses. Finally we use this dataset to propose a first solution to the problem by training a neural network to predict a "semantic score" from a top view image and camera pose. The views predicted to have higher scores are then shown to provide better clustering results than fixed top-down views.

This paper proposes the first step towards a novel unified framework for the analysis of perturbations occurring in nuclear reactors in both Time and Frequency domain. The identification of type and source of such perturbations is fundamental for monitoring core reactors and guarantee safety even while running at nominal conditions. A 3D Convolutional Neural Network (3D-CNN) was employed to analyse perturbations happening in the frequency domain, such as the alteration of an absorber of variable strength or propagating perturbation. Recurrent neural networks (RNN), specifically Long Short-Term Memory (LSTM) was used to study signal sequences related to perturbations induced in the time domain, including the vibrations of fuel assemblies and the fluctuation of thermalhydraulic parameters at the inlet of the reactor coolant loops. 512-dimensional representations were extracted from the 3D-CNN and LSTM architectures, and used as input to a fused multi-sigmoid classification layer to recognise the perturbation type. If the perturbation is frequency domain related, a separate fully-connected layer utilises said representations to regress the coordinates of its source. The results showed that perturbation type can be recognised with high accuracy in both domains, and frequency domain scenario sources can be localised with high precision.

Toribash is a MuJoCo-like environment of two humanoid character fighting each other hand-to-hand, controlled by changing states of body joints. Competitive nature of Toribash lends itself to two-agent experiments, and active player-base can be used for human baselines. This white paper describes the environment with its pros, cons and limitations as well experimentally show ToriLLE's applicability as a learning environment by successfully training reinforcement learning agents that improved over time. The code is available at https: //github.com/Miffyli/ToriLLE.

Although there is an emerging trend towards generating embeddings for primarily unstructured data, and recently for structured data, there is not yet any systematic suite for measuring the quality of embeddings. This deficiency is further sensed with respect to embeddings generated for structured data because there are no concrete evaluation metrics measuring the quality of encoded structure as well as semantic patterns in the embedding space. In this paper, we introduce a framework containing three distinct tasks concerned with the individual aspects of ontological concepts: (i) the categorization aspect, (ii) the hierarchical aspect, and (iii) the relational aspect. Then, in the scope of each task, a number of intrinsic metrics are proposed for evaluating the quality of the embeddings. Furthermore, w.r.t. this framework multiple experimental studies were run to compare the quality of the available embedding models. Employing this framework in future research can reduce misjudgment and provide greater insight about quality comparisons of embeddings for ontological concepts.

Interactive reinforcement learning (IRL) extends traditional reinforcement learning (RL) by allowing an agent to interact with parent-like trainers during a task. In this paper, we present an IRL approach using dynamic audio-visual input in terms of vocal commands and hand gestures as feedback. Our architecture integrates multi-modal information to provide robust commands from multiple sensory cues along with a confidence value indicating the trustworthiness of the feedback. The integration process also considers the case in which the two modalities convey incongruent information. Additionally, we modulate the influence of sensory-driven feedback in the IRL task using goal-oriented knowledge in terms of contextual affordances. We implement a neural network architecture to predict the effect of performed actions with different objects to avoid failed-states, i.e., states from which it is not possible to accomplish the task. In our experimental setup, we explore the interplay of multimodal feedback and task-specific affordances in a robot cleaning scenario. We compare the learning performance of the agent under four different conditions: traditional RL, multi-modal IRL, and each of these two setups with the use of contextual affordances. Our experiments show that the best performance is obtained by using audio-visual feedback with affordancemodulated IRL. The obtained results demonstrate the importance of multi-modal sensory processing integrated with goal-oriented knowledge in IRL tasks.

Generating large quantities of quality labeled data in medical imaging is very time consuming and expensive. The performance of supervised algorithms for various tasks on imaging has improved drastically over the years, however the availability of data to train these algorithms have become one of the main bottlenecks for implementation. To address this, we propose a semi-supervised learning method where pseudo-negative labels from unlabeled data are used to further refine the performance of a pulmonary nodule detection network in chest radiographs. After training with the proposed network, the false positive rate was reduced to 0.1266 from 0.4864 while maintaining sensitivity at 0.89.

It's a plot straight out of science fiction: Bad guys dispose of an unlucky security guard, scoop out one of the guy's (or gal's) eyeballs, and hold it up to an iris scanner, fooling it into disarming a security system. As it turns out, post-mortem eyes can be used for biometric identification hours or even days after death, studies show. But if researchers at Warsaw University of Technology in Poland have their way, that might not be the case for much longer. In a paper ("Presentation Attack Detection for Cadaver Irises") published on the preprint server Arxiv.org, the team proposed a neural network that can tell the difference between living irises and dead ones with 99 percent accuracy. "With increasing importance that biometric authentication gains in our daily lives, fears are increasingly common among users, regarding the possibility of unauthorized access to our data, identity, or assets after our demise," the researchers wrote.

One of the things that we learned from the DARPA Robotics Challenge is that it's useful for robots to have legs to walk, but it's even more useful for robots to be versatile and adaptable, with multimodal locomotion capabilities that they can deploy depending on the situation. At the DRC, we saw all kinds of different designs, but one of the more unique approaches came from the University of Bonn, in Germany, with their robot Momaro. Momaro used a "centaur" design, with four legs that had wheels on the bottom (like a wheeled quadruped) coupled to a humanoid upper torso with a head and arms. It was the top-ranked European robot in the DRC, completing an almost perfect run in just 34 minutes. We've since been wondering whether the centaur design would inspire other disaster robots, and now we know the answer is yes.