I am really proud to announce that our latest effort on showing the advantages of our analog/digital #neuromorphic spiking neural network chips in solving complex biomedical applications has just been published here: https://rdcu.be/ef5N0
The demonstrates that *small* *highly variable* and *low accuracy* #SNNs can indeed be useful, without having to resort to #backprop in large-scale #DNNs! 😉
'Densely Connected G-invariant Deep Neural Networks with Signed Permutation Representations', by Devanshu Agrawal, James Ostrowski.
http://jmlr.org/papers/v24/23-0294.html
#representations #dnns #dnn
This presentation at DeepSec 2023 explores the use of Artificial Intelligence Large Language Models for social engineering campaigns.
Training DNNs Resilient to Adversarial and Random Bit-Flips by Learning Quantization Ranges
Beyond Distribution Shift: Spurious Features Through the Lens of Training Dynamics
Compressors such as #gzip + #kNN (k-nearest-neighbor i.e. your grandparents' #classifier) beats the living daylights of Deep neural networks (#DNNs) in sentence classification.
H/t @lgessler
Without any training parameters, this non-parametric, easy and lightweight (no #GPU) method achieves results that are competitive with non-pretrained deep learning methods on six in-distribution datasets.It even outperforms BERT on all five OOD datasets.
'Integrating Random Effects in Deep Neural Networks', by Giora Simchoni, Saharon Rosset.