We presented a framework for evolutionary robotics in both simulation world and the real world by integrating Gym, Gazebo, and ROS. The interface provided by Gym allows users and researchers to focus on the learners, instead of wasting time on the s…

With Contextual BERT, we presented novel ways of conditioning the BERT model. The strong performance on a real-world use-case provides evidence for the superiority of using a global state to inject context into the Transformer-based architecture. Ou…

High and low frequency components present in an image play a vital role when they are processed by deep learning models. Several recent research works have also highlighted that CNN models are highly sensitive towards high and low-frequency componen…

Converting sequential data into a format which implicitly encodes temporal information as spatial information is an effective way of generating samples of such data as whole pieces. Here, we explored this approach for melody generation of fixed-leng…

In this paper, we propose a simple single-loop algorithm for nonconvex min-max problems (1.1). For an important family of problems (1.2), the algorithm is even more efficient due to the dual error bound, and it is well-suited for problems in large-s…

In this paper, we propose a theoretical framework using structural causal models (SCMs) to analyze and obtain insights on prediction performance of several popular domain adaptation methods. First, we show that under the assumption of anticausal pre…

This work was supported by the Engineering and Physical Sciences Research Council (EPSRC), UK. This work was also supported by European Union’s Horizon 2020 Research and Innovation Program H2020 ICT 2016-2017 (as an initiative of the Photonics Publi…

In this paper, we presented a method for analyzing metastable chemical reaction systems via feature learning on directed networks using random walk sampling. We have shown how this method may be used to identify transition states of Markov jump proc…

Many open problems in machine learning are intrinsically related to causality, however, the use of causal analysis in machine learning is still in its early stage. Within a general reinforcement learning setting, we consider the problem of building …

The main conclusion of our research is that temporal context matters. All the papers came to the results that their models, which use more than one frame to detect objects outperform the models with similar baseline architecture which only proceed o…

We have introduced a scalable nonparametric learning algorithm to estimate the interaction kernel from discrete data with performance guarantee. The algorithm learns the kernel on a data-adaptive hypothesis space by least squares with regularization…