_index.org

Likelihood Free Bayesian Optimization

Last edited: October 10, 2026

High level motivation: black-box optimization

\begin{equation} x^{*} \in \arg\max_{x \in S} f\qty(x) \end{equation}

where \(x\) is some candidate design, \(f\) the objective, and generally we can only see evaluations of \(y = f\qty(x) + \varepsilon\).

Baysian Optimization

A basic method is to fit a Gaussian process; given:

\begin{equation} D_{t-1} = \qty {\qty(x_{i}, y_{i})}_{i=1}^{t-1} \end{equation}

The historical measurement we fit some kind of parametrized probability distribution, say, a guassian process:

\begin{equation} f\qty(x) \sim \mathcal{GP} \qty(u_{t}, k_{t}) \end{equation}

query selection

Last edited: October 10, 2026

Each query presents one item and records a binary response (did they like it?).

  • Rasch: a person parameter \(\alpha\)
  • Frozen encoder: reward-heads weights \(w\)

We don’t want to present too many queries to train this.

Fisher Information

Fisher recognizes fixed \(u\), and have expectation under different measures Y

New User, Known Object

Suppose you want to know the information gain from understanding \(\alpha\) the user parameter, conditioned on some known item you tested \(\beta\). Consider some random variable \(Y\) under some kind of Rasch model:

Shielding Safe AI Systems

Last edited: October 10, 2026

Safe Leraning Problems

How do we make decision making agents abide by safety requirement at runtime? Given \(T\) and spec \(\phi\), find optimal \(\pi^{*}\) to solve \(T\) such that \(\pi^{*} \text{ SAT } \phi\) using some notion of satisfaction.

  • What is the notion of safety?
  • what are the requirements in temporal logic?
  • How do we enforce compliance?
  • How do we prove compliance?

why not just apply formal methods to AI?

  • environments change…
  • notion of safety for stochastic systems are random
  • neural networks are not easily probable since they are approximations of what we are trying to model

MASA Library

MASA uses shielding as the key technique for ensuring safety.

Stanford UG Courses Index

Last edited: October 10, 2026

Stanford UG Y1, Aut

Stanford UG Y1, Win

Stanford UG Y1, Spr

Stanford UG Y2, Aut

Stanford UG Y2, Win

Stanford UG Y2, Spr

Stanford UG Y3, Aut

Stanford UG Y3, Win

Stanford UG Y3, Spr

Stanford GR Y1, Aut

Stanford Talks

DateTopicPresenterLink
<2023-09-20 Wed>UG Research ProgramBrian ThomasStanford UG Research Program
<2023-09-28 Thu>Bld an Ecosystem, Not MonolithColin RaffelBuild a System
<2023-10-05 Thu>Training Helpful CHatbotsNazeen RajaniTraining Helpful Chatbots
<2023-10-26 Thu>AI Intepretability for BioGasper BegusAI Intepretability
<2023-11-02 Thu>PT Transformers on Long SeqsMike LewisPretraining Long Transformers
<2023-11-07 Tue>Transformers!A. VaswaniTransformers
<2023-11-09 Thu>Towards Interactive AgentsJessy LinInteractive Agent
<2023-11-16 Thu>Dissociating Language and ThoughtAnna IvanovaDissociating Language and Thought
<2024-01-11 Thu>Language AgentsKarthik NarasimhanLanguage Agents with Karthik
<2024-02-01 Thu>Pretraining Data
<2024-02-08 Thu>value alignmentBeen KimLM Alignment
<2024-02-15 Thu>model editingPeter HaseKnowledge Editing
<2024-07-18 Thu>Knowledge Localization
<2024-11-11 Mon>PresentationsSydney KatzPresentations
<2025-01-06 Mon>Video Generation with Learned PriorMeenakshi SarkarPriors
<2025-01-06 Mon>Theoretical Drone ControlSliding Mode UAV Control
<2025-01-09 Thu>VLM to AgentsTao YuVLM to Agents
<2025-01-13 Mon>Social RLNatasha JaquesSocial Reinforcement Learning
<2025-02-10 Mon>Model Predictive Control + PromptingGabriel MaherLLM MPC
<2025-03-03 Mon>Planning for Learning
<2025-03-06 Thu>Theorem ProvingSelf-Play Conjection Generalization
<2025-04-10 Thu>Safety for TrucksSafety for Autonomous Trucking
<2025-08-04 Mon>Collaborate Multiagent DMCollaborative Multiagent DM
<2025-09-22 Mon>AI Safety TalksAI Safety Annual Meeting
<2025-10-02 Thu>Pretraining under infinite computeLimited Samples and Infinite Compute
<2025-10-06 Mon>Mel KrusniakDecisions.jl
<2025-10-11 Sat>SISL Flash TalksSISL Talks
<2025-10-16 Thu>Predicting Scaling Performance
<2025-12-08 Mon>mixed-autonomy traffic with LLMSmixed-autonomy traffic with LLMs
<2026-01-05 Mon>AI Incidents PolicyAI Incidents Policy
<2026-01-12 Mon>Reliable RLReliable RL
<2026-01-15 Thu>Words to ConceptsWords to Concepts
Zen’s Defense
<2026-03-30 Mon>multi-agent LLMMulti-Agent LLMs
<2026-04-27 Mon>Alex’s Defense
<2026-09-28 Mon>Yi Chengen ai safety model and datasets
<2026-09-28 Mon>Anastasia KoslovaFrom Security to Trustworthy AI
<2026-10-12 Mon>Problems with AI
<2026-10-12 Mon>Likelihood Free Bayesian Optimization

Contacts

Talk Contacts