My new journal paper “UNSUPERVISED ROTATION FACTORIZATION IN RESTRICTED BOLTZMANN MACHINES” is now available at HTTPS://IEEEXPLORE.IEEE.ORG/DOCUMENT/8870198. This paper shows a new method to learn rotation-invariant features on shallow neural network, without the necessity to use data augmentation. The paper shows mathematical and experimental proofs of its effectiveness.
From January 18th to February 2nd, I spent two weeks in Debre Zeyt, Ethiopia, as part of my task for the BBSRC GCF project I am currently working on (further details at http://chickpearoots.org). We provided in-situ support to build a rhizobox imaging station, replicating the setup we had built in Edinburgh. It has been an incredible experience, from the professional and personal point of view. The take-away thought was: international cooperation is important in collaborative research.
The 7th of June I was invited by Dr. Devis Tuia to give a talk regarding my research activities. I have ~30 mins presentation on Machine Learning for Plant Phenotyping, showing my latest findings on rotation-invariance (paper recently submitted). I would like to thank Devis and his incredible research group for the warm welcoming and inspiring discussions we had last week.
The 4th CVPPP workshop will be held in conjunction with BMVC 2018, Newcastle (UK) on the 6th September 2018. More info here!