Generative dialogue models currently suffer from a number of problems which standard maximum likelihood training does not address. They tend to produce generations that (i) rely too much on copying from the context, (ii) contain repetitions within utterances, (iii) overuse frequent words, and (iv) at a deeper level, contain logical flaws. In this work we show how all of these problems can be addressed by extending the recently introduced unlikelihood loss (Welleck et al., 2019a) to these cases. We show that appropriate loss functions which regularize generated outputs to match human distributions are effective for the first three issues. For the last important general issue, we show applying unlikelihood to collected data of what a model should not do is effective for improving logical consistency, potentially paving the way to generative models with greater reasoning ability. We demonstrate the efficacy of our approach across several dialogue tasks.

We investigate the impact of search strategies in neural dialogue modeling. We first compare two standard search algorithms, greedy and beam search, as well as our newly proposed iterative beam search which produces a more diverse set of candidate responses. We evaluate these strategies in realistic full conversations with humans and propose a modelbased Bayesian calibration to address annotator bias. These conversations are analyzed using two automatic metrics: log-probabilities assigned by the model and utterance diversity. Our experiments reveal that better search algorithms lead to higher rated conversations. However, finding the optimal selection mechanism to choose from a more diverse set of candidates is still an open question.

In this work we release our extensible and easily configurable neural network training software. It provides a rich set of functional layers with a particular focus on efficient training of recurrent neural network topologies on multiple GPUs. The source of the software package is public and freely available for academic research purposes and can be used as a framework or as a standalone tool which supports a flexible configuration. The software allows to train state-of-the-art deep bidirectional long short-term memory (LSTM) models on both one dimensional data like speech or two dimensional data like handwritten text and was used to develop successful submission systems in several evaluation campaigns.

No abstract

The article describes an elaboration of the X-in-the-loop (XiL) testing environment for a thermal management system (TMS) intended for the traction electric drive of an electric vehicle, which has each of its wheels driven by an in-wheel motor. The TMS features the individual thermal regulation of each electric drive using a hydraulic layout with parallel pipelines and electrohydraulic pumps embedded into them. The XiL system is intended as a tool for studying and developing the TMS design and controls. It consists of the virtual part and the physical part. The former simulates the vehicle operating in a driving cycle with the heat power dissipated by the electric drive components, which entails the change in their temperature regimes. The physical part includes the TMS itself consisting of a radiator, pipelines, and pumps. The physical part also features devices intended for simulation of the electric drive components in terms of their thermal and hydraulic behaviors, as well as devices that simulate airflow induced by the vehicle motion. Bilateral, real-time interactions are established between the two said parts combining them into a cohesive system, which models the studied electric vehicle and its components. The article gives a description of a laboratory setup, which implements the XiL environment including the mathematical models, hardware devices, as well as the control loops that establish the interaction of those components. An example of using this system in a driving cycle test shows the interaction between its parts and operation of the TMS in conditions simulated in both virtual and physical domains. The results constitute calculated and measured quantities including vehicle speed, operating parameters of the electric drives, coolant and air flow rates, and temperatures of the system components.