Speech Production Features for Deep Neural Network Acoustic Modeling - 2015
Explore speech production knowledge for improved DNN acoustic modeling in ASR. Learn about vocal tract measurements and phonological embeddings to enhance recognition accuracy.
Introduction to Artificial Intelligence: Deep Neural Networks - Lecture 19
Explore deep neural networks, edge detection, and reinforcement learning in AI, with applications to games and image processing.
Introduction to Deep Neural Networks - Lecture 21
Explore deep neural networks, their architecture, and applications in AI. Learn key concepts and techniques for implementing advanced machine learning models.
Interpreting Deep Neural Networks Towards Trustworthiness
Explore methods for interpreting neural networks, including contextual decomposition and adaptive wavelet distillation, to enhance model trustworthiness and performance.
Geometry, Topology and Discrete Symmetries Revealed by Deep Neural Networks
Explore deep neural networks' ability to reveal geometry, topology, and discrete symmetries in manifolds. Learn about universal approximators and their applications in inverse problems and group actions.
Advanced Computational Imaging - Chassez le Naturel, Il Revient au Galop
Explorez les avancées en imagerie computationnelle, en mettant l'accent sur la stabilité et la fiabilité des réseaux neuronaux profonds pour la reconstruction d'images biomédicales.
Next-Generation Networks for Machine Learning
Explore next-gen networks for machine learning, focusing on optical fabrics and scheduling techniques to accelerate distributed DNN training workloads and overcome datacenter network bottlenecks.
A View in Old and New Machine Learning
Explore old and new machine learning approaches, comparing hierarchical probabilistic models with deep neural networks and examining their unique capabilities and potential combinations.
The Mathematical Universe Behind Deep Neural Networks - Rothschild Lecture
Explore the mathematical foundations of deep neural networks, covering functional analysis, harmonic analysis, complex analysis, approximation theory, and more with Professor Helmut Bölcskei.
Solving the Many-Electron Schrödinger Equation with Deep Neural Networks
Explore deep neural networks for solving complex quantum mechanical equations in many-electron systems, advancing computational physics and materials science.
Data-Driven Modeling of Unknown Systems with Deep Neural Networks
Explore data-driven modeling of unknown systems using deep neural networks. Learn to discover and approximate evolution operators for predictive analysis across various problem domains.
Polynomial Time Cryptanalytic Extraction of Neural Network Models
Explore polynomial-time techniques for extracting neural network parameters, improving upon previous methods and demonstrating practical efficiency on a CIFAR10 classifier.
Learning Manifold-Structured Data Using Deep Neural Networks - Theory and Applications
Explore deep neural networks' ability to learn geometric data structures using multi-chart latent spaces and Chart Auto-Encoders, with theoretical foundations and practical applications.
Visual Features for Context-Aware Speech Recognition - 2016
Explore visual features for context-aware speech recognition, improving transcription of consumer-generated multimedia content by adapting acoustic and language models to detected objects and scenes.
Intro to Artificial Intelligence - Deep Neural Networks - Lecture 20
Explore deep neural networks, their architecture, and applications in AI. Learn key concepts and techniques for implementing advanced machine learning models.
Implementation of Backtracking Line Search in Deep Neural Networks - Theory and Practice
Explore backtracking line search implementation in deep neural networks, covering theoretical foundations and practical applications for optimizing model performance.
Splitting the Difference on Adversarial Training - A Method for Robust Neural Networks
Discover a novel approach to adversarial training that enhances neural network robustness while maintaining optimal accuracy by splitting data classes into clean and adversarial categories.
DNN-GP: Diagnosing and Mitigating Model's Faults Using Latent Concepts
Explore an innovative fault diagnosis tool for Deep Neural Networks that interprets latent concepts to detect and mitigate various model vulnerabilities, from adversarial attacks to concept drift.
Tossing in the Dark: Practical Bit-Flipping on Gray-box Deep Neural Networks for Runtime Trojan Injection
Explore practical bit-flipping techniques for runtime Trojan injection in deep neural networks, focusing on gray-box scenarios and innovative approaches to bypass memory protection systems.
Yes, One-Bit-Flip Matters! Universal DNN Model Inference Depletion with Runtime Code Fault Injection
Explore FrameFlip, a groundbreaking attack method that depletes DNN model inference through single-bit fault injections, demonstrating universal effectiveness across major machine learning frameworks.