In 2017-18, only six per cent of Australian soft plastic waste was recycled. The rest added to the growing mountain of plastic in landfills around the country. The biggest problem is the lack of an automatic solution to sort soft plastic waste from co-mingled recycling. Vucetic, an Engineers Australia Fellow, explained that this is because soft plastics like bread bags and cling wrap get tangled in machinery, causing equipment failures and contaminating other waste streams. Sydney-based recycling provider iQRenew invited Vucetic's team at the University of Sydney's Centre for IoT and Telecommunications to see the problem first hand, and potentially help them automate their processes.

Gaussian Conditional Random Fields (GCRF) are atype of structured regression model that incorporatesmultiple predictors and multiple graphs. This isachieved by defining quadratic term feature functions inGaussian canonical form which makes the conditionallog-likelihood function convex and hence allows findingthe optimal parameters by learning from data. In thiswork, the parameter space for the GCRF model is extendedto facilitate joint modelling of positive and negativeinfluences. This is achieved by restricting the modelto a single graph and formulating linear bounds on convexitywith respect to the models parameters. In addition,our formulation for the model using one networkallows calculating gradients much faster than alternativeimplementations. Lastly, we extend the model onestep farther and incorporate a bias term into our linkweight. This bias is solved as part of the convex optimization.Benefits of the proposed model in terms ofimproved accuracy and speed are characterized on severalsynthetic graphs with 2 million links as well as on ahospital admissions prediction task represented as a humandisease-symptom similarity network correspondingto more than 35 million hospitalization records inCalifornia over 9 years.

Conditional probabilistic graphical models provide a powerful Thus, a particular interest of this paper is long-term forecasting framework for structured regression in spatiotemporal on non-static networks with continuous target variables datasets with complex correlation patterns. It has been (structured regression) and proper uncertainty propagation shown that models utilizing underlying correlation patterns estimate in such evolving networks. This is motivated (structured models) can significantly improve predictive accuracy by climate modeling of long-term precipitation prediction in as compared to models not utilizing such information spatiotemporal weather station networks, as well as prediction (Radosavljevic, Vucetic, and Obradovic 2010; 2014; of different disease trends in temporal disease-disease Ristovski et al. 2013; Wytock and Kolter 2013; Stojanovic networks.