43 citations · 97 across the 18 of their papers we have counts for
5 papers · 1 filter
Self-Supervised Transformers for fMRI representation
Itzik Malkiel, Gony Rosenman, Lior Wolf +1
We present TFF, which is a Transformer framework for the analysis of functional Magnetic Resonance Imaging (fMRI) data. TFF employs a two-phase training approach. First, self-super…
Caption Enriched Samples for Improving Hateful Memes Detection
Efrat Blaier, Itzik Malkiel, Lior Wolf
The recently introduced hateful meme challenge demonstrates the difficulty of determining whether a meme is hateful or not. Specifically, both unimodal language models and multimod…
GAM: Explainable Visual Similarity and Classification via Gradient Activation Maps
Oren Barkan, Omri Armstrong, Amir Hertz +4
We present Gradient Activation Maps (GAM) - a machinery for explaining predictions made by visual similarity and classification models. By gleaning localized gradient and activatio…
Self-Supervised Document Similarity Ranking via Contextualized Language Models and Hierarchical Inference
Dvir Ginzburg, Itzik Malkiel, Oren Barkan +2
We present a novel model for the problem of ranking a collection of documents according to their semantic similarity to a source (query) document. While the problem of document-to-…
Adaptive Gradient Balancing for Undersampled MRI Reconstruction and Image-to-Image Translation
Itzik Malkiel, Sangtae Ahn, Valentina Taviani +3
Recent accelerated MRI reconstruction models have used Deep Neural Networks (DNNs) to reconstruct relatively high-quality images from highly undersampled k-space data, enabling muc…