➕Probabilistic-Artificial-Intelligence

There is huge literature on planning. We will attack this problem from the view of probabilistic artificial intelligence. In this case we focus on continuous, fully observed with non-linear transitions, an environment often used for robotics. It’s called Model Predictive Control (MPC). \[...\] Moreover, modeling uncertainty in our model of the environment can be extremely useful in deciding where to explore. Learning a model can therefore help to dramatically reduce the sample complexity over model-free techniques. ...

These algorithms are good for scaling state spaces, but not actions spaces. The Gradient Idea Recall Temporal difference learning and Q-Learning, two model free policy evaluation techniques explored in Tabular Reinforcement Learning. A simple parametrization 🟩 The idea here is to parametrize the value estimation function so that similar inputs gets similar values akin to Parametric Modeling estimation we have done in the other courses. In this manner, we don’t need to explicitly explore every single state in the state space. ...

This note extends the content Markov Processes in this specific context. Standard notions Explore-exploit dilemma 🟩 We have seen something similar also in Active Learning when we tried to model if we wanted to look elsewhere or go for the maximum value we have found. The dilemma under analysis is the explore-exploit dilemma: whether if we should just go for the best solution we have found at the moment, or look for a better one. This also has implications in many other fields, also in normal human life there are a lot of balances in these terms. ...

$$ p(\theta \mid x_{1:n}, y_{1:n}) = \frac{1}{z} p(y_{1:n} \mid \theta, x_{1:n}) p(\theta \mid x_{1:n}) \approx q(\theta \mid \lambda) $$For Bayesian Linear Regression we had high dimensional Gaussians which made the inference closed form, in general this is not true, so we need some kinds of approximation. Laplace approximation Introduction to the Idea 🟩 $$ \psi(\theta) \approx \hat{\psi}(\theta) = \psi(\hat{\theta}) + (\theta-\hat{\theta} ) ^{T} \nabla \psi(\hat{\theta}) + \frac{1}{2} (\theta-\hat{\theta} ) ^{T} H_{\psi}(\hat{\theta})(\theta-\hat{\theta} ) = \psi(\hat{\theta}) + \frac{1}{2} (\theta-\hat{\theta} ) ^{T} H_{\psi}(\hat{\theta})(\theta-\hat{\theta} ) $$ We simplified the term on the first order because we are considering the mode, so the gradient should be zero for the stationary point. ...

Introduzione alle catene di Markov La proprietà di Markov Una sequenza di variabili aleatorie $X_{1}, X_{2}, X_{3}, \dots$ gode della proprietà di Markov se vale: $$ P(X_{n}| X_{n - 1}, X_{n - 2}, \dots, X_{1}) = P(X_{n}|X_{n-1}) $$ Ossia posso scordarmi tutta la storia precedente, mi interessa solamente lo stato precedente per sapere la probabilità attuale. Da un punto di vista filosofico/fisico, ha senso perché mi sta dicendo che posso predire lo stato successivo se ho una conoscenza (completa, (lo dico io completo, originariamente non esiste)) del presente. ...

The maximum entropy principle is one of the most important guiding motives in artificial artificial intelligence. Its roots emerge from a long tradition of probabilistic inference that goes back to Laplace and Occam’s Razor, i.e. the principle of parsimony. Let’s start with a simple example taken from Andreas Kraus’s Lecture notes in the ETH course of Probabilistic Artificial Intelligence: Consider a criminal trial with three suspects, A, B, and C. The collected evidence shows that suspect C can not have committed the crime, however it does not yield any information about sus- pects A and B. Clearly, any distribution respecting the data must assign zero probability of having committed the crime to suspect C. However, any distribution interpolating between (1, 0, 0) and (0, 1, 0) respects the data. The principle of indifference suggests that the desired distribution is $(\frac{1}{2}, \frac{1}{2}, 0)$, and indeed, any alterna- tive distribution seems unreasonable. ...

The main difference between reinforcement learning and other machine learning, pattern inference methods is that reinforcement learning takes the concept of actions into its core: models developed in this field can be actively developed to have an effect in its environment, while other methods are mainly used to summarize interesting data or generating sort of reports. Reinforcement learning (RL) is an interdisciplinary area of machine learning and optimal control concerned with how an intelligent agent ought to take actions in a dynamic environment in order to maximize the cumulative reward. ~Wikipedia page. ...

This note will mainly attempt to summarize the introduction of some intuitive notions of probability used in common sense human reasoning. Most of what is said here is available here (Jaynes 2003). Three intuitive notions of probability Jaynes presents some forms of inference that are not possible in classical first order or propositional logic, yet they are frequent in human common sense reasoning. Let’s present some rules and some examples along them: ...