Game theory has traditionally had a relatively limited view of risk based on how a player's expected reward is impacted by the uncertainty of the actions of other players. Recently, a new game-theoretic approach provides a more holistic view of risk also considering the reward-variance. However, these variance-based approaches measure variance of the reward on both the upside and downside. In many domains, such as finance, downside risk only is of key importance, as this represents the potential losses associated with a decision. In contrast, large upside "risk" (e.g. profits) are not an issue. To address this restrictive view of risk, we propose a novel solution concept, downside risk aware equilibria (DRAE) based on lower partial moments. DRAE restricts downside risk, while placing no restrictions on upside risk, and additionally, models higher-order risk preferences. We demonstrate the applicability of DRAE on several games, successfully finding equilibria which balance downside risk with expected reward, and prove the existence and optimality of this equilibria.

We propose a novel composite reward function for a reinforcement learning (RL) trading agent that explicitly balances return and risk by combining four differentiable components--annualized return, downside risk, differential return, and the Treynor ratio. Unlike traditional single-metric objectives (e.g., Sharpe or cumulative return), which can encourage reward hacking or over-optimization of one aspect of trading, our formulation is inherently modular and weighted w

Long-term planning and production scheduling are among the most critical tasks in the area of mining. The goal is to extract valuable ore from an orebody in a sequence that takes into account many mining and precedence constraints in a way that is economically efficient [1]. This is an important real-world optimisation problem that has been studied in the literature over many years. This includes mixed integer programming approaches based on block scheduling [2, 3]. Each block in a block model (a discretised spatial approximation of the mineral deposit) has a given estimated value based on the metal grade and the excavation cost. Other heuristic techniques include dealing with specific characteristics such as uncertainties of the problem [4-6]. Different software products that offer a variety of approaches for mine planning and extraction sequences are available [7, 8]. Evolutionary computation techniques have successfully been applied in the area of mining, in particular to large scale optimisation problems such as the cost efficient extraction of ore [9, 10], the ore processing and blending problem [11-15], and the large-scale open pit mine scheduling problem [16, 17]. Particle swarm algorithms were utilised to solve the capacity constrained open pit mining problem [18] and the transportation and layout problem of locating a crushing station in an open-pit mine [19].

Existing work on risk-sensitive reinforcement learning - both for symmetric and downside risk measures - has typically used direct Monte-Carlo estimation of policy gradients. While this approach yields unbiased gradient estimates, it also suffers from high variance and decreased sample efficiency compared to temporal-difference methods. In this paper, we study prediction and control with aversion to downside risk which we gauge by the lower partial moment of the return. We introduce a new Bellman equation that upper bounds the lower partial moment, circumventing its non-linearity. We prove that this proxy for the lower partial moment is a contraction, and provide intuition into the stability of the algorithm by variance decomposition. This allows sample-efficient, on-line estimation of partial moments. For risk-sensitive control, we instantiate Reward Constrained Policy Optimization, a recent actor-critic method for finding constrained policies, with our proxy for the lower partial moment. We extend the method to use natural policy gradients and demonstrate the effectiveness of our approach on three benchmark problems for risk-sensitive reinforcement learning.

Sixty-four percent of the more than 1,200 senior security executives from around the world, whom we surveyed for the 2018 Thales Data Threat Report (DTR), believe artificial intelligence (AI) "increases data security by recognizing and alerting on attacks," while 43% believe AI "results in increased threats due to use as a hacking tool." On the one hand, security executives can use AI and its subset technology, machine learning (ML), to enhance digital security. For example, they can use AI to look for unusual security events and find those needles in a haystack faster. They can also use AI to detect malware and more. In this context, AI/ML is no more than training a system to learn to find things faster.

One Big Question: How do we manage the downside risks of AI? Darren Quick If Hollywood is to be believed, the development of super-intelligent AI will spell the end of civilization as we know it and spark an unwinnable war between man and machine. Be the first to comment by clicking the button below.

If Hollywood is to be believed, the development of super-intelligent AI will spell the end of civilization as we know it and spark an unwinnable war between man and machine. It doesn't make for nearly as exciting entertainment, but artificial intelligence also offers tremendous upside, from the potential to deliver customized education to everyone, to improving disease diagnosis and treatment and eradicating poverty. Although AI researchers are focused these beneficial outcomes, the dystopian vision portrayed in so much science fiction is also a real possibility. At the recent Singularity University (SU) New Zealand Summit we talked with Neil Jacobstein, the former president and current chair of the Artificial Intelligence and Robotics Track at SU, about how the outcomes feared by so many can be avoided.

IBM (NYSE:IBM) has a lot of moving parts, and a vociferous crowd of critics. In the process of analysis, it's easy to be overwhelmed by complexity, or sidetracked into refuting mindless attacks by the ill-informed. In the interest of simplicity, this article focuses on hard evidence of the company's progress in exploiting the developing market for Artificial Intelligence, Machine Learning, or Cognitive Computing. The quarterly and annual financial results provide segment information, to include year over year revenue growth and pre-tax margins. Looking at 2Q 2016, Cognitive Solutions at 4.4% is the only segment showing growth, and at 27.5% the second highest (after Global Financing) in pre-tax margins.