Adversarial machine learning is a growing field that focuses on testing and enhancing the resilience of machine learning (ML) systems through adversarial examples. These examples are crafted by subtly altering data to deceive the models into making incorrect predictions. Deep generative models (DGMs) have shown significant promise in generating such adversarial examples, especially in computer… →

Monocular depth estimation (MDE) plays an important role in various applications, including image and video editing, scene reconstruction, novel view synthesis, and robotic navigation. However, this task poses significant challenges due to the inherent scale distance ambiguity, making it ill-posed. Learning-based methods should utilize robust semantic knowledge to achieve accurate results and overcome this limitation.… →

With the rapid advancement of technology, surpassing human abilities in tasks like image classification and language processing, evaluating the energy impact of ML is essential. Historically, ML projects prioritized accuracy over energy efficiency, contributing to increased energy consumption. Green software engineering, highlighted by Gartner as a key trend for 2024, focuses on addressing this issue.… →

Training a Large CNN for Image Classification:Researchers developed a large CNN to classify 1.2 million high-resolution images from the ImageNet LSVRC-2010 contest, spanning 1,000 categories. The model, which contains 60 million parameters and 650,000 neurons, achieved impressive results, with top-1 and top-5 error rates of 37.5% and 17.0%, respectively—significantly outperforming previous methods. The architecture comprises… →

In the significantly advancing fields of neuroscience and Artificial Intelligence (AI), the goal of comprehending and modeling human cognition has resulted in the creation of sophisticated models that make an effort to mimic the intricate workings of the brain. The tensor brain, a computer framework created to mimic how the brain interprets and stores information,… →

Nvidia unveiled its latest large language model (LLM) offering, the Llama-3.1-Nemotron-51B. Based on Meta’s Llama-3.1-70B, this model has been fine-tuned using advanced Neural Architecture Search (NAS) techniques, resulting in a breakthrough in both performance and efficiency. Designed to fit on a single Nvidia H100 GPU, the model significantly reduces memory consumption, computational complexity, and costs… →