Simone Marullo

It has been amazing to have been back to a physical @IJCAIconf #IJCAI2022 together with @enricomeloni_ai and @MarulloSimone‼️ "Stochastic Coherence Over Attention Trajectory For Continuous Learning In Video Streams" proceedings: ijcai.org/proceedings/20…

Exploitation of language neural representations for the retrieval and ranking of crossword clues and answers presented by @AZugarini ceur-ws.org/Vol-3033/paper… #clicit2021 #NLProc @AILC_NLP

#IJCAI2022 "Stochastic Coherence Over Attention Trajectory For Continuous Learning In Video Streams" by @TiezziMatteo, @MarulloSimone, Lapo Faggi, @enricomeloni_ai, @aleitteb and @StefanoMelacci. Soon @IJCAIconf ‼️ Technical report: arxiv.org/abs/2204.12193 7/7

Thanks to #NFT, noisy information can be discarded. We introduced some noise in #CIFAR10 dataset. #NFT foster the standard training process (CT) of a ResNet18. A work by @MarulloSimone , @TiezziMatteo , @KleineBottleM , @StefanoMelacci. Soon #AAAI2022 ! arxiv.org/abs/2112.09968

In Neural Friendly Training (#NFT 🐵), we show that major improvements can be obtained by introducing a temporary auxiliary neural model implementing a structured and shared transformation function - to spot regularities in the data alteration process arxiv.org/abs/2112.09968

Friendly Training (FT) alter the training data to facilitate the learning process of a neural classifier. Transformations discard the parts of information that are too complex to be handled by the network with the current weights. arxiv.org/abs/2106.10974

In Adversarial Machine Learning, Neural models alter data samples, aimed at fooling a classifier. Why not be a friend? "Being Friends Instead of Adversaries: Deep Networks Learn from Data Simplified by Other Networks" #AAAI22 @RealAAAI! arxiv.org/abs/2112.09968 👇🧵

Friendly Training: Neural Networks Can Adapt Data To Make Learning Easier deepai.org/publication/fr… by Simone Marullo et al. including @TiezziMatteo #DeepLearning #NeuralNetwork