Shastri, HetviHetviShastriBatra, NipunNipunBatra2025-08-312025-08-312021-11-17[9781450391146]10.1145/3486611.34866522-s2.0-85120934165http://repository.iitgn.ac.in/handle/IITG2025/25215Non-intrusive load monitoring (NILM) involves separating the household aggregate energy consumption into constituent appliances. In 2014, a toolkit called NILMTK was released towards making NILM reproducible. Subsequently, in 2019, an improved version called NILMTK-contrib, focused on experiments and ease of adding new algorithms was released. Since then, there have been significant advances in neural networks for various applications, and in the NILM domain. In this paper, we implement five recent neural network architectures for NILM in NILMTK-contrib and benchmark against existing algorithms. Further, in this paper, we also implement a dataset parser for a publicly available dataset called IDEAL containing 255 homes with 39 homes having appliance data. We find that the new algorithms are comparable or better than the state-of-the-art over a subset of the appliances.falsedatasets | energy disaggregation | neural networks | NILMConference Paper172-17517 November 20219cpConference Proceeding11