We present methods for optimizing portfolios, asset allocations, and trading systems based on direct reinforcement (DR). In this approach, investment decision-making is viewed as a stochastic control problem, and strategies are discovered directly. We present an adaptive algorithm called recurrent reinforcement learning (RRL) for discovering investment policies. The need to build forecasting mo...

In this paper, decentralized reinforcement learning is applied to a control problem with a multidimensional action space. We propose a decentralized reinforcement learning architecture for a mobile robot, where the individual components of the commanded velocity vector are learned in parallel by separate agents. We empirically demonstrate that the decentralized architecture outperforms its cent...

The use of differential reinforcement of other behavior (DRO) has decreased, at least partially due to the development of less effortful alternative behavioral interventions (e.g., noncontingent reinforcement; Vollmer, Iwata, Zarcone, Smith, & Mazaleski, 1993). The effort associated with DRO contingencies may be lessened by incorporating self-monitoring components in which clients are responsib...

This article reviews the practical value of conceptual attempts to specify the circumstances of reinforcement ahead of time. Improvements are traced from the transituational-reinforcer approach of Meehl (1950), through the probability-differential model of Premack (1959, 1965), to the response deprivation and disequilibrium approach (Timberlake, 1980, 1984; Timberlake & Allison, 1974). The appl...

Rats increased eating that produced access to a running-wheel or increased running that produced access to food. depending on which response was potentially deprived. relative to baseline. by the scheduled ratio of responding. Under both schedules. instrumental responding significantly exceeded appropriate baselines of the noncontingent effects of the schedule. The results contradicted the hypo...

We propose to train trading systems by optimizing financial objective functions via reinforcement learning. The performance functions that we consider are profit or wealth, the Sharpe ratio and our recently proposed differential Sharpe ratio for online learning. In Moody & Wu (1997), we presented empirical results that demonstrate the advantages of reinforcement learning relative to supervised ...

Behavioral momentum theory provides a quantitative account of how reinforcers experienced within a discriminative stimulus context govern the persistence of behavior that occurs in that context. The theory suggests that all reinforcers obtained in the presence of a discriminative stimulus increase resistance to change, regardless of whether those reinforcers are contingent on the target behavio...