Neural Network – Buonaiuto@Work

SpikingBrain: a revolutionary brain-inspired Chatgpt made in China

Massimo2025-09-180357 views

The Chinese SpikingBrain is a new family of brain-inspired large language models that reimagines how AI can process information more efficiently. SpikingBrain models adopt a biological principle: neurons remain idle until an event triggers them to fire. This event-driven design reduces unnecessary computation, cuts energy use, and enables faster responses. SpikingBrain achieves over 100× speedup in “time to first token” for sequences up to 4 million tokens. Energy consumption drops by 97% compared to traditional LLMs.

Markov Chains, MDPs, and Memory-Augmented MDPs: The Mathematical Core of Agentic AI

Massimo2025-09-010147 views

Markov Chains, Markov Decision Processes (MDP), and Memory-augmented MDPs (M-MDP) form the mathematical backbone of decision-making under uncertainty. While Markov Chains capture stochastic dynamics, MDPs extend them with actions and rewards. Yet, real-world tasks demand memory—this is where M-MDPs shine. By embedding structured memory into the agent’s state, M-MDPs enable agentic AI systems to reason, plan, and adapt across long horizons. This blog post explores the mathematics, technicalities, and the disruptive role of M-MDPs in modern AI architectures.

Generative AI in the Automotive Industry

Massimo2024-05-090437 views

Generative AI is revolutionizing the automotive industry, enhancing design, supply chain management, and predictive maintenance. By optimizing designs and customizing features, AI is enabling rapid prototyping and improving operational efficiency. AI also boosts supply chain resilience by predicting disruptions and automating quality control, ensuring high standards. The technology’s role in developing autonomous vehicles through extensive scenario testing highlights its transformative impact.

Find cancer with AI: a closer look at CT scan analysis with Self-Supervised Learning (SSL)

Massimo2024-04-080425 views

In the ever-evolving battle against cancer, the integration of Self-Supervised Learning (SSL) with CT scan analysis emerges as a beacon of hope, illuminating new pathways for early and accurate diagnosis. SSL, a sophisticated facet of machine learning, thrives on the challenge of unlabeled data, teaching AI models to navigate through vast informational landscapes to uncover hidden patterns indicative of cancer. This pioneering approach not only promises to enhance the precision of cancer detection but also to streamline the operational efficiency of healthcare diagnostics. By leveraging the untapped potential of SSL, we stand on the cusp of revolutionizing how we identify and combat cancer, making strides towards a future where accurate diagnosis is both faster and more accessible.

Interleaving algorithm for optimization of neural networks with self-learning perceptrons

Massimo2022-01-010405 views

Exploring the Efficiency of Interleaving Algorithms in Neural Network Optimization, this study introduces a novel application of team draft interleaving, diverging from traditional A/B testing methods. By simulating a sports team selection process, this approach enhances compound selection from a dataset. Highlighting its utility in artificial intelligence, particularly in self-learning perceptrons, the method enables perceptrons to adapt activation functions dynamically. This preemptive adjustment, facilitated by interleaving, marks a significant departure from conventional error backpropagation, demonstrating potential for more responsive learning mechanisms in neural networks.

The (Un)reliability of Saliency methods – Google Research

Massimo2021-07-160485 views

In the exploration of deep model interpretation, saliency methods emerge as a popular technique for evaluating feature importance. They assign importance scores to input features, indicating their utility in model performance. High scores suggest significant performance degradation in their absence. However, investigations, such as those by Google Research, reveal the inherent unreliability of these methods. The crux of the issue lies in their sensitivity to non-influential factors and failure to maintain input invariance, leading to potentially misleading attributions. This challenges the effectiveness of saliency methods in providing accurate explanations of deep learning behaviors.

Research on Climate Change with statistics methods

Massimo2016-08-290945 views

Analyzing vast datasets from GBIF and meteorological organizations, we unveil how climate change impacts plant behavior and distribution worldwide.