Notes

Trusted Execution Environments and the Secure Computation Trust Spectrum Motivation & Foundational Context How do you verify that a computation was performed correctly — and privately — when you don’t control the machine? This question sits at the intersection of Cryptography, Hardware Security, Mechanism Design, and Multi-Agent Systems. Different answers trade off trust assumptions, performance, and expressiveness, forming a trust spectrum that has deep consequences for agent coordination, blockchain protocols, and verifiable AI. ...

Bayesian Inverse Reinforcement Learning (BIRL) Background: The IRL Problem Setup Inverse Reinforcement Learning sits at the intersection of Reinforcement Learning, Bayesian Inference, and Imitation Learning. The core question is: given observed expert behavior, what reward function is the expert optimizing? This is strictly harder than forward RL, and the key challenge is that the problem is fundamentally ill-posed. The IRL Problem Definition IRL: Given an MDP without a reward function $\mathcal{M} \setminus R = \langle \mathcal{S}, \mathcal{A}, T, \gamma \rangle$ and a set of expert demonstrations $\mathcal{D} = \{(s_i, a_i)\}$, recover the reward function $R : \mathcal{S} \times \mathcal{A} \to \mathbb{R}$ the expert is (approximately) optimizing. ...

an Here we attempt to answer what is memory, how is it stored and retrieved. Human Memory Definition of Memory Memory is a process by which information is: Encoded Stored Retrieved The brain has different types of memories, and certain brain regions are specialized for this task. Ebbinghaus Curves Other experiments destroy parts of the cortex and correlate this with recall. Types of memory We have the memory pipeline: ...

We have a prior $p(\text{model})$, we have a posterior $p(\text{model} \mid \text{data})$, a likelihood $p(\text{data} \mid \text{model})$ and $p(\text{data})$ is called the evidence. Classical Linear regression $$ y = w^{T}x + \varepsilon $$ Where $\varepsilon \sim \mathcal{N}(0, \sigma_{n}^{2}I)$ and it’s the irreducible noise, an error that cannot be eliminated by any model in the model class, this is also called aleatoric uncertainty. One could write this as follows: $y \sim \mathcal{N}(w^{T}x, \sigma^{2}_{n}I)$ and it’s the exact same thing as the previous, so if we look for the MLE estimate now we get ...

Introduction to Communication Games We start by defining the fundamental problem: strategic information transmission between agents where information is asymmetric. The Communication Problem Information Asymmetry: One player (Sender) knows something the other (Receiver) does not. Strategic Interaction: Both players maximize their own utility, which may or may not align. Signaling: The act of sending a message to convey (or hide) information. Credibility: The central property determining if communication influences actions. ...

Industrial Organization: Scale vs. Flexibility Let’s consider a fundamental model for production in a competitive economy. Here we are interested to predict the survival probability of a firm, and how resources are allocated between capital investment and labor adaptability. For simplicity, we will assume two distinct agents: the Consolidated Corporation ($C$) and the Networked SME ($S$). Types of variables Economists will say that parameters for this model, i.e., variables that are fixed for the analysis, are exogenous variables. In this setting, the primary exogenous variable is Market Volatility ($\sigma$). The variables of interest, or endogenous variables, are the Firm Size ($Q$) and the Reaction Time ($\tau$). ...

Introduzione (idea principale) In breve: essence card Giallo = Prodotto. Metafora staffetta-rugby Con altri metodi si fanno produzioni stile staffetta, ossia un membro sta fermo, finché non ha il testimone e poi si uccide correndo… Il metodo più utile ispirato a scrum è rugby, che tutti si muovo insieme collaborando. Un po’ di tutto è fatto durante lo sprint Cicli di base (3) Planning: in cui vengono scelti i task da eseguire durante questo sprint, solitamente questo viene preso da un subset dei task descritti dal product owner. Execution: questo è abbastanza chiaro, si sviluppa. Retrospective and review: in cui vengono identificati i problemi che sono stati incontrati durante lo sviluppo, e modi possibili per risolverli. ! 500 Lo sprint (3) Una cosa molto importante che aiuterà di gran lunga lo sviluppo è la costanza che Si scelgono ...

Hopfield networks are a type of recurrent neural network that can store and retrieve patterns. They are particularly useful for associative memory tasks, where the network can recall a stored pattern given a noisy or partial input. The Hopfield Network Model $$ \Theta = \sum_{i = 1}^{s} \left[ x_{t}x_{t}^{T} - \boldsymbol{I}_{n} \right] $$ One can prove that this matrix is symmetric, and that it is irreflexive (meaning $\forall j: \Theta_{jj} = 0$). ...

Introduction: a neuron I am lazy, so I’m skipping the introduction for this set of notes. Look at Andrew Ng’s Coursera course for this part (here are the notes). Historical paper is (Rosenblatt 1958). One can view a perceptron to be a Log Linear Models with the temperature of the softmax that goes to 0 (so that it is an argmax). Trained with a stochastic gradient descent with a batch of 1 (this is called the perceptron update rule, see The Perceptron Model). ...

Dependable systems Introduzione Possiamo individuare alcune proprietà dei sistemi distribuiti. Però non siamo riusciti a renderli logicamente validi. Sono ancora un pò misti di linguaggio naturale e della sua ambiguità! Comunque possiamo ridurci per guardare quanto un sistema sia affidabile a guardare poche sue caratteristiche precise. Caratteristiche fondamentali (4) Queste proprietà sono pensate naturalmente caratterizzanti dei sistemi. In particolare dovrebbero essere tutti misurabili. Availability Che risponde nell’istante in cui fai una richiesta. ...