In recent years, generative models have shown remarkable capabilities across diverse fields, including images, videos, language, and decision-making. By applying powerful generative models such as flow-based models to reinforcement learning, we can effectively model complex multi-modal action distributions and achieve superior robotic control in continuous action spaces, surpassing the limitations of single-modal action distributions with traditional Gaussian-based policies. Previous methods usually adopt the generative models as behavior models to fit state-conditioned action distributions from datasets, with policy optimization conducted separately through additional policies using value-based sample weighting or gradient-based updates. However, this separation prevents the simultaneous optimization of multi-modal distribution fitting and policy improvement, ultimately hindering the training of models and degrading the performance. To address this issue, we propose Decision Flow, a unified framework that integrates multi-modal action distribution modeling and policy optimization. Specifically, our method formulates the action generation procedure of flow-based models as a flow decision-making process, where each action generation step corresponds to one flow decision. Consequently, our method seamlessly optimizes the flow policy while capturing multi-modal action distributions. We provide rigorous proofs of Decision Flow and validate the effectiveness through extensive experiments across dozens of offline RL environments. Compared with established offline RL baselines, the results demonstrate that our method achieves or matches the SOTA performance.

In this manuscript we introduce a novel Decision Flow (DF) framework for sampling from a target distribution while incorporating additional guidance from a prior sampler. DF can be viewed as an AI driven algorithmic reincarnation of the Markov Decision Process (MDP) approach in Stochastic Optimal Control. It extends the continuous space, continuous time path Integral Diffusion sampling technique to discrete time and space, while also generalizing the Generative Flow Network framework. In its most basic form, an explicit, Neural Network (NN) free formulation, DF leverages the linear solvability of the the underlying MDP to adjust the transition probabilities of the prior sampler. The resulting Markov Process is expressed as a convolution of the reverse time Green's function of the prior sampling with the target distribution. We illustrate the DF framework through an example of sampling from the Ising model, discuss potential NN based extensions, and outline how DF can enhance guided sampling across various applications.

Technology research and consulting firm Gartner, Inc. identified Decision Intelligence as a top strategic trend for 2022. Analyst Dr. Pieter J. den Hamer begins his detailed report with this call to action: "To deal with unprecedented levels of business complexity and uncertainty, organizations must make accurate and highly contextualized decisions more quickly. IT leaders must create capabilities to rapidly compose and recompose transparent decision flows." Few would argue with the idea that businesses have an ever greater need to make better, faster decisions. However, the report has an underlying insight that portends the fundamental disruption of this most important leadership activity.

In recent years, medical resources in China have been in a state of short supply. Doctors from the top hospitals in modern cities have to face tedious consultation or surgical work every day. Meanwhile, some doctors also need to do scientific research related work for the development of medicine. However, in doctors' daily work, a large part of the work is cumbersome but easy to use IT systems to improve efficiency, such as the management of medical terms, the collection of medical knowledge, and so on. The organization of medical knowledge is mostly for daily diagnosis and treatment in accordance with certain gold standards or guidelines, and many of them are based on some established rules to make inference.

With self-driving cars cruising around, robots doing backflips and helping each other open doors, computers learning how to play GO in a few days and then beating experts who spent their lives mastering the game, we are definitely witnessing an exciting era in human history. Like Enova's CTO John Higginson said in his blog post, as an analytics and technology company we want to use and even seek to extend these technologies to invent the future, but for credit. That's exactly why our executive team picked'advancing our machine learning capabilities' as one of our strategic initiatives this year. Usually when people see these amazing advancements in technology the first thing they think about is how machines are taking over our jobs. In the lending industry, however, the takeover has already happened.