Traffic signal control (TSC) is still one of the most significant and challenging research problems in the transportation field. Reinforcement learning (RL) has achieved great success in TSC but suffers from critically high learning costs in practical applications due to the excessive trial-and-error learning process. Offline RL is a promising method to reduce learning costs whereas the data distribution shift issue is still up in the air. To this end, in this paper, we formulate TSC as a sequence modeling problem with a sequence of Markov decision process described by states, actions, and rewards from the traffic environment. A novel framework, namely TransformerLight, is introduced, which does not aim to fit into value functions by averaging all possible returns, but produces the best possible actions using a gated Transformer. Additionally, the learning process of TransformerLight is much more stable by replacing the residual connections with gated transformer blocks due to a dynami
This paper considers an NP-hard network topology design (NTD) problem called NTD-CB/R. A key challenge when solving the bi-objective optimisation problem is to simultaneously minimise cost while maximising bandwidth. This paper aims to generate the best set of non-dominated feasible topologies, known as the Pareto Optimal Set (POS). It formally defines a dynamic programming (DP) formulation for NTD-CB/R. Then, it proposes two alternative Lagrange relaxations to compute a weight for each link. The paper proposes a DP approach, called DPCB/R-LP, to generate POS with maximum weight. Extensive simulations on hundreds of networks that contain up to 299 paths show that DPCB/R-LP can generate 70.4% of the optimal POS while using only up to 984 paths and 27.06 CPU seconds. Overall-Pareto-spread (OR), DPCB/R-LP produces 94.4% of POS with OS = 1, measured against the optimal POS. Finally, all generated POS’s with largest bcr, significantly higher than 88% obtained by existing methods.
EXCLUSIVE: The Hot Springs Documentary Film Festival – the country’s longest running all-documentary festival – is preparing to launch an innovative new Filmmaker Forum. The inaugural event, with funding provided by the Corporation for Public Broadcasting, will run from October 9-10 as part of the 32nd edition of the festival in Hot Springs, Ark. The festival …
VisuAlgo was conceptualised in 2011 by Dr Steven Halim as a tool to help his students better understand data structures and algorithms, by allowing them to learn the basics on their own and at their own pace. Together with his students from the National University of Singapore, a series of visualizations were developed and consolidated, from simple sorting algorithms to complex graph data structures. Though specifically designed for the use of NUS students taking various data structure and algorithm classes (CS1010/equivalent, CS2040/equivalent, CS3230, CS3233, and CS4234), as advocators of online learning, we hope that curious minds around the world will find these visualizations useful as well.