Adversarial Training (AT) has become arguably the state-of-the-art algorithm for extracting robust features. However, researchers recently notice that AT suffers from severe robust overfitting problems, particularly after learning rate (LR) decay. In this paper, we explain this phenomenon by viewing adversarial training as a dynamic minimax game between the model trainer and the attacker. Specifically, we analyze how LR decay breaks the balance between the minimax game by empowering the trainer with a stronger memorization ability, and show such imbalance induces robust overfitting as a result of memorizing non-robust features. We validate this understanding with extensive experiments, and provide a holistic view of robust overfitting from the dynamics of both the two game players. This understanding further inspires us to alleviate robust overfitting by rebalancing the two players by either regularizing the trainer's capacity or improving the attack strength.

Payment channel networks (PCNs) are a layer-2 blockchain scalability solution, with its main entity, the payment channel, enabling transactions between pairs of nodes "off-chain," thus reducing the burden on the layer-1 network. Nodes with multiple channels can serve as relays for multihop payments by providing their liquidity and withholding part of the payment amount as a fee. Relay nodes might after a while end up with one or more unbalanced channels, and thus need to trigger a rebalancing operation. In this paper, we study how a relay node can maximize its profits from fees by using the rebalancing method of submarine swaps. We introduce a stochastic model to capture the dynamics of a relay node observing random transaction arrivals and performing occasional rebalancing operations, and express the system evolution as a Markov Decision Process. We formulate the problem of the maximization of the node's fortune over time over all rebalancing policies, and approximate the optimal solution by designing a Deep Reinforcement Learning (DRL)-based rebalancing policy. We build a discrete event simulator of the system and use it to demonstrate the DRL policy's superior performance under most conditions by conducting a comparative study of different policies and parameterizations. Our work is the first to introduce DRL for liquidity management in the complex world of PCNs.

Artificial intelligence, or AI, is everywhere right now. In truth, the fundamentals of AI and machine learning have been around for a long time. The first primitive form of AI was an automated checkers bot which was created by Cristopher Strachey from the University of Manchester, England, back in 1951. It's come a long way since then, and we're starting to see a large number of high profile use cases for the technology being thrust into the mainstream. Some of the hottest applications of AI include the development of autonomous vehicles, facial recognition software, virtual assistants like Amazon's AMZN Alexa and Apple's AAPL Siri and a huge array of industrial applications in all industries from farming to gaming to healthcare.

Artificial intelligence, or AI, is everywhere right now. In truth, the fundamentals of AI and machine learning have been around for a long time. The first primitive form of AI was an automated checkers bot which was created by Cristopher Strachey from the University of Manchester, England, back in 1951. It's come a long way since then, and we're starting to see a large number of high profile use cases for the technology being thrust into the mainstream. Some of the hottest applications of AI include the development of autonomous vehicles, facial recognition software, virtual assistants like Amazon's Alexa and Apple's Siri and a huge array of industrial applications in all industries from farming to gaming to healthcare.

Offline reinforcement learning (RL) is challenged by the distributional shift between learning policies and datasets. To address this problem, existing works mainly focus on designing sophisticated algorithms to explicitly or implicitly constrain the learned policy to be close to the behavior policy. The constraint applies not only to well-performing actions but also to inferior ones, which limits the performance upper bound of the learned policy. Instead of aligning the densities of two distributions, aligning the supports gives a relaxed constraint while still being able to avoid out-of-distribution actions. Therefore, we propose a simple yet effective method to boost offline RL algorithms based on the observation that resampling a dataset keeps the distribution support unchanged. More specifically, we construct a better behavior policy by resampling each transition in an old dataset according to its episodic return. We dub our method ReD (Return-based Data Rebalance), which can be implemented with less than 10 lines of code change and adds negligible running time. Extensive experiments demonstrate that ReD is effective at boosting offline RL performance and orthogonal to decoupling strategies in long-tailed classification. New state-of-the-arts are achieved on the D4RL benchmark.

Over the decades, the Markowitz framework has been used extensively in portfolio analysis though it puts too much emphasis on the analysis of the market uncertainty rather than on the trend prediction. While generative adversarial network (GAN), conditional GAN (CGAN), and autoencoding CGAN (ACGAN) have been explored to generate financial time series and extract features that can help portfolio analysis. The limitation of the CGAN or ACGAN framework stands in putting too much emphasis on generating series and finding the internal trends of the series rather than predicting the future trends. In this paper, we introduce a hybrid approach on conditional GAN based on deep generative models that learns the internal trend of historical data while modeling market uncertainty and future trends. We evaluate the model on several real-world datasets from both the US and Europe markets, and show that the proposed HybridCGAN and HybridACGAN models lead to better portfolio allocation compared to the existing Markowitz, CGAN, and ACGAN approaches.

If, as William Gibson once wrote, "the future is here, it's just not very evenly distributed", then artificial intelligence (AI) has long topped the list of technologies waiting to be better distributed. Historically, it's been the preserve of enterprise-scale organisations, able to gamble on long-term, resource-intensive processes which held no guarantee of paying off. Recent years, however, have seen AI, machine learning (ML) and robotic process automation (RPA) become...

Online portfolio selection (PS) has been extensively studied in artificial intelligence and machine learning communities in recent years. An important practical issue of online PS is transaction cost, which is unavoidable and nontrivial in real financial trading markets. Most existing strategies, such as universal portfolio (UP) based strategies, often rebalance their target portfolio vectors at every investment period, and thus the total transaction cost increases rapidly and the final cumulative wealth degrades severely. To overcome the limitation, in this paper we investigate new investment strategies that rebalances its portfolio only at some selected instants. Specifically, we design a novel on-line PS strategy named semi-universal portfolio (SUP) strategy under transaction cost, which attempts to avoid rebalancing when the transaction cost outweighs the benefit of trading. We show that the proposed SUP strategy is universal and has an upper bound on the regret. We present an efficient implementation of the strategy based on non-uniform random walks and online factor graph algorithms. Empirical simulation on real historical markets show that SUP can overcome the drawback of existing UP based transaction cost aware algorithms and achieve significantly better performance. Furthermore, SUP has a polynomial complexity in the number of stocks and thus is efficient and scalable in practice.

Albert Bifet University of W aikato, New Zealand LTCI, T el ecom ParisT ech, France abifet@waikato.ac.nz Abstract --Nowadays, every device connected to the Internet generates an ever-growing stream of data (formally, unbounded). Machine Learning on unbounded data streams is a grand challenge due to its resource constraints. In fact, standard machine learning techniques are not able to deal with data whose statistics is subject to gradual or sudden changes without any warning. Massive Online Analysis (MOA) is the collective name, as well as a software library, for new learners that are able to manage data streams. In this paper, we present a research study on streaming rebalancing. Indeed, data streams can be imbalanced as static data, but there is not a method to rebalance them incrementally, one element at a time. For this reason we propose a new streaming approach able to rebalance data streams online. Our new methodology is evaluated against some synthetically generated datasets using prequential evaluation in order to demonstrate that it outperforms the existing approaches.

There is a troubling convergence of trends across the cybersecurity landscape that I have been watching closely. If not addressed, I suspect they could wreak deeper levels of damage and volatility than any we have already seen. Cybercriminals are taking advantage of the expanding attack surfaces being created by digital transformation, the extraordinary ease and accessibility of malware as both off-the-shelf product and emerging profit driver, and the fact that IT teams are often so overwhelmed managing change that they simply don't have the resources necessary to keep systems appropriately patched and hardened. When we take a closer look, the challenges are stark, but the solution doesn't require a genius, if we approach it wisely and methodically. After all, genius is 1 percent inspiration and 99 percent perspiration.