Image translation with convolutional neural networks has recently been used as an approach to multimodal change detection. Existing approaches train the networks by exploiting supervised information of the change areas, which, however, is not always available. A main challenge in the unsupervised problem setting is to avoid that change pixels affect the learning of the translation function. We propose two new network architectures trained with loss functions weighted by priors that reduce the impact of change pixels on the learning objective. The change prior is derived in an unsupervised fashion from relational pixel information captured by domain-specific affinity matrices. Specifically, we use the vertex degrees associated with an absolute affinity difference matrix and demonstrate their utility in combination with cycle consistency and adversarial training. The proposed neural networks are compared with state-of-the-art algorithms. Experiments conducted on two real datasets show the effectiveness of our methodology.

In this paper, we introduce DeepQuarantine (DQ), a cloud technology to detect and quarantine potential spam messages. Spam attacks are becoming more diverse and can potentially be harmful to email users. Despite the high quality and performance of spam filtering systems, detection of a spam campaign can take some time. Unfortunately, in this case some unwanted messages get delivered to users. To solve this problem, we created DQ, which detects potential spam and keeps it in a special Quarantine folder for a while. The time gained allows us to double-check the messages to improve the reliability of the anti-spam solution. Due to high precision of the technology, most of the quarantined mail is spam, which allows clients to use email without delay. Our solution is based on applying Convolutional Neural Networks on MIME headers to extract deep features from large-scale historical data. We evaluated the proposed method on real-world data and showed that DQ enhances the quality of spam detection.

In the context of supervised statistical learning, it is typically assumed that the training set comes from the same distribution that draws the test samples. When this is not the case, the behavior of the learned model is unpredictable and becomes dependent upon the degree of similarity between the distribution of the training set and the distribution of the test set. One of the research topics that investigates this scenario is referred to as domain adaptation. Deep neural networks brought dramatic advances in pattern recognition and that is why there have been many attempts to provide good domain adaptation algorithms for these models. Here we take a different avenue and approach the problem from an incremental point of view, where the model is adapted to the new domain iteratively. We make use of an existing unsupervised domain-adaptation algorithm to identify the target samples on which there is greater confidence about their true label. The output of the model is analyzed in different ways to determine the candidate samples. The selected set is then added to the source training set by considering the labels provided by the network as ground truth, and the process is repeated until all target samples are labelled. Our results report a clear improvement with respect to the non-incremental case in several datasets, also outperforming other state-of-the-art domain adaptation algorithms.

Compared to supervised learning, semi-supervised learning reduces the dependence of deep learning on a large number of labeled samples. In this work, we use a small number of labeled samples and perform data augmentation on unlabeled samples to achieve image classification. Our method constrains all samples to the predefined evenly-distributed class centroids (PEDCC) by the corresponding loss function. Specifically, the PEDCC-Loss for labeled samples, and the maximum mean discrepancy loss for unlabeled samples are used to make the feature distribution closer to the distribution of PEDCC. Our method ensures that the inter-class distance is large and the intra-class distance is small enough to make the classification boundaries between different classes clearer. Meanwhile, for unlabeled samples, we also use KL divergence to constrain the consistency of the network predictions between unlabeled and augmented samples. Our semi-supervised learning method achieves the state-of-the-art results, with 4000 labeled samples on CIFAR10 and 1000 labeled samples on SVHN, and the accuracy is 95.10% and 97.58% respectively.

When people learn mathematical patterns or sequences, they are able to identify the concepts (or rules) underlying those patterns. Having learned the underlying concepts, humans are also able to generalize those concepts to other numbers, so far as to even identify previously unseen combinations of those rules. Current state-of-the art RNN architectures like LSTMs perform well in predicting successive elements of sequential data, but require vast amounts of training examples. Even with extensive data, these models struggle to generalize concepts. From our behavioral study, we also found that humans are able to disregard noise and identify the underlying rules generating the corrupted sequences. We therefore propose a Bayesian model that captures these human-like learning capabilities to predict next number in a given sequence, better than traditional LSTMs.

Recent advancements in Artificial Intelligence (AI) have brought new capabilities to behavioural analysis (UEBA) for cyber-security consisting in the detection of hostile action based on the unusual nature of events observed on the Information System.In our previous work (presented at C\&ESAR 2018 and FIC 2019), we have associated deep neural networks auto-encoders for anomaly detection and graph-based events correlation to address major limitations in UEBA systems. This resulted in reduced false positive and false negative rates, improved alert explainability, while maintaining real-time performances and scalability. However, we did not address the natural evolution of behaviours through time, also known as concept drift. To maintain effective detection capabilities, an anomaly-based detection system must be continually trained, which opens a door to an adversary that can conduct the so-called "frog-boiling" attack by progressively distilling unnoticed attack traces inside the behavioural models until the complete attack is considered normal. In this paper, we present a solution to effectively mitigate this attack by improving the detection process and efficiently leveraging human expertise. We also present preliminary work on adversarial AI conducting deception attack, which, in term, will be used to help assess and improve the defense system. These defensive and offensive AI implement joint, continual and active learning, in a step that is necessary in assessing, validating and certifying AI-based defensive solutions.

Estimating temporal patterns in travel times along road segments in urban settings is of central importance to traffic engineers and city planners. In this work, we propose a methodology to leverage coarse-grained and aggregated travel time data to estimate the street-level travel times of a given metropolitan area. Our main focus is to estimate travel times along the arterial road segments where relevant data are often unavailable. The central idea of our approach is to leverage easy-to-obtain, aggregated data sets with broad spatial coverage, such as the data published by Uber Movement, as the fabric over which other expensive, fine-grained datasets, such as loop counter and probe data, can be overlaid. Our proposed methodology uses a graph representation of the road network and combines several techniques such as graph-based routing, trip sampling, graph sparsification, and least-squares optimization to estimate the street-level travel times. Using sampled trips and weighted shortest-path routing, we iteratively solve constrained least-squares problems to obtain the travel time estimates. We demonstrate our method on the Los Angeles metropolitan-area street network, where aggregated travel time data is available for trips between traffic analysis zones. Additionally, we present techniques to scale our approach via a novel graph pseudo-sparsification technique.

A reflex is a simple closed loop control approach which tries to minimise an error but fails to do so because it will always react too late. An adaptive algorithm can use this error to learn a forward model with the help of predictive cues. For example a driver learns to improve their steering by looking ahead to avoid steering in the last minute. In order to process complex cues such as the road ahead deep learning is a natural choice. However, this is usually only achieved indirectly by employing deep reinforcement learning having a discrete state space. Here, we show how this can be directly achieved by embedding deep learning into a closed loop system and preserving its continuous processing. We show specifically how error back-propagation can be achieved in z-space and in general how gradient based approaches can be analysed in such closed loop scenarios. The performance of this learning paradigm is demonstrated using a line-follower both in simulation and on a real robot that show very fast and continuous learning.

Click here to read the full story with my Friend Link! When I tell people that they should consider applying deep learning methods to their data, a common initial response I get is: (1) "I am not working with big enough data," and (2) "I do not have access to enough computational resources to train deep learning models." I believe these assumptions come from large companies (e.g. Google) that often like to show off by conducting research on large datasets, such as ImageNet which contains over a million pictures, and by using a large amount of GPUs. That's great for these companies, but from my impression, the average deep learning practitioner is not working with such large datasets and does not have access to such large computational resources.

This means that the test favors "program synthesis," the subfield of AI that involves generating programs that satisfy high-level specifications. This approach is in contrast with current trends in AI, which are inclined toward creating programs that are optimized for a limited set of tasks (e.g., playing a single game). In his experiments with ARC, Chollet has found that humans can fully solve ARC tests.