Google TechTalks Courses

Revisiting Nearest Neighbors from a Sparse Signal Approximation View

Explore nearest neighbor methods from a sparse signal approximation perspective, uncovering improved neighborhood definitions and their applications in graph-based signal processing and machine learning.

• YouTube
• 50 minutes
• On-Demand
• Free Video

Engineering Block-based Languages - Day 1 Session 9

Explore block-based programming languages, focusing on KOGI and its applications in education and language development.

• YouTube
• 20 minutes
• On-Demand
• Free Video

A Constant Factor Prophet Inequality for Online Combinatorial Auctions

Explore constant factor prophet inequalities for online combinatorial auctions with subadditive valuations. Learn about the Mirror Lemma and Kakutani's fixed point theorem application.

• YouTube
• 59 minutes
• On-Demand
• Free Video

Open Problems in Mechanistic Interpretability: A Whirlwind Tour

Explore key concepts and open problems in mechanistic interpretability of neural networks, including techniques for reverse engineering learned algorithms and improving AI safety.

• YouTube
• 55 minutes
• On-Demand
• Free Video

Online Prediction in Sub-linear Space

Explore innovative algorithms for online learning with expert advice, focusing on sub-linear memory solutions and their implications for sequential decision-making.

• YouTube
• 1 hour
• On-Demand
• Free Video

Accelerating Transformers via Kernel Density Estimation - Google TechTalk

Explore efficient Transformer acceleration using kernel density estimation. Learn innovative techniques for handling long sequences and improving performance in deep learning models.

• YouTube
• 53 minutes
• On-Demand
• Free Video

Differentially Private Synthetic Data Generation via Foundation Model APIs

Explore generating high-quality differentially private synthetic data using foundation model APIs, overcoming challenges and achieving state-of-the-art results without accessing model weights.

• YouTube
• 56 minutes
• On-Demand
• Free Video

Foundation Models and Fair Use - Legal and Ethical Implications

Explore legal and ethical implications of training foundation models on copyrighted material. Analyze fair use doctrine, potential risks, and technical mitigations for model development and deployment.

• YouTube
• 57 minutes
• On-Demand
• Free Video

Differentially Private Diffusion Models Generate Useful Synthetic Images

Explore privacy-preserving synthetic image generation using diffusion models. Learn techniques for fine-tuning pre-trained models to achieve state-of-the-art results on CIFAR-10 and Camelyon17 datasets.

• YouTube
• 41 minutes
• On-Demand
• Free Video

Randomized Approach for Tight Privacy Accounting in Differential Privacy

Explore a novel differential privacy paradigm for improved privacy accounting in compositions, featuring Monte Carlo-based verification and enhanced utility-privacy tradeoffs in machine learning.

• YouTube
• 46 minutes
• On-Demand
• Free Video

Improving the Privacy-Utility Tradeoff in Differentially Private Machine Learning with Public Data

Explore techniques to enhance privacy-utility balance in differentially private machine learning using public data, focusing on DP-RAFT and DOPE-SGD algorithms for improved model accuracy.

• YouTube
• 38 minutes
• On-Demand
• Free Video

EIFFeL - Ensuring Integrity for Federated Learning

Explore secure federated learning with EIFFeL, a framework ensuring update privacy and integrity while detecting and removing malformed data to prevent model poisoning.

• YouTube
• 31 minutes
• On-Demand
• Free Video

Almost Tight Error Bounds on Differentially Private Continual Counting

Explore novel mechanisms for differentially private continual counting, analyzing error bounds and applications in private optimization. Gain insights into matrix mechanisms and factorization techniques.

• YouTube
• 56 minutes
• On-Demand
• Free Video

Differentially Private Online-to-Batch Conversion for Stochastic Optimization

Explore differentially private stochastic optimization techniques, comparing simple preprocessing and custom algorithms. Learn about a novel online-to-batch conversion method for optimal private convergence rates.

• YouTube
• 50 minutes
• On-Demand
• Free Video

Differentially Private Diffusion Models

Explore privacy-preserving machine learning with differentially private diffusion models, optimizing design for sensitive data and enhancing performance through public pre-training.

• YouTube
• 45 minutes
• On-Demand
• Free Video