Deep Amortized Inference for Probabilistic Programs

By Daniel Ritchie, Paul Horsfall, Noah D. Goodman. 2016

87 citations recorded by the source corpus.

Probabilistic programming languages (PPLs) are a powerful modeling tool, able to represent any computable probability distribution. Unfortunately, probabilistic program inference is often intractable, and existing PPLs mostly rely on expensive, approximate sampling-based methods. To alleviate this problem, one could try to learn from past inferences, so that future inferences run faster. This strategy is known as amortized inference; it has recently been applied to Bayesian networks and deep generative models. This paper proposes a system for amortized inference in PPLs. In our system, amortization comes in the form of a parameterized guide program. Guide programs have similar structure to the original program, but can have richer data flow, including neural network components. These networks can be optimized so that the guide approximately samples from the posterior distribution defined by the original program. We present a flexible interface for defining guide programs and a stochastic gradient-based scheme for optimizing guide parameters, as well as some preliminary results on automatically deriving guide programs. We explore in detail the common machine learning pattern in which a 'local' model is specified by 'global' random values and used to generate independent observed data points; this gives rise to amortized local inference supporting global model learning.

Subfields: Artificial Intelligence

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

Citations in the corpus

Relational inductive biases, deep learning, and graph networks (2018, methodology): “…numerous recent examples of principled hybrids of structure-based methods and deep learning (e.g., Reed and De Freitas, 2016; Garnelo et al., 2016; Ritchie et al., 2016; Wu et al., 2017; Denil et al., 2017; Hudson and Manning, 2018), we see great promise in synthesizing new techniques by drawing…”
Learning Disentangled Representations with Semi-Supervised Deep Generative Models (2017, background): “This approach allows us also amortise inference [7, 23, 29, 34], simultaneously learning a network that performs approximate inference over representations at the same time we learn the unknown factors of the model itself.”
Automatic differentiation in machine learning: a survey (2015, background): “, 2013) techniques based on deep learning (Le et al., 2017; Ritchie et al., 2016) work by training neural networks for performing approximate inference in generative models defined as probabilistic programs (Gordon et al.”
Deep Probabilistic Programming (2017, background): “Concurrent with our work, WebPPL features amortized inference (Ritchie et al., 2016).”
Meta-learning of Sequential Strategies (2019, background): “14 Realizable is when the minimizer f ∗ of the Monte-Carlo loss is in F . i.e. the optimal sequential prediction in which the Bayesian update is automatically amortized (or pre-computed) [Ritchie et al., 2016].”
Transformers Can Do Bayesian Inference (2021, background): “, 2018), which was previously adapted to general probabilistic programs (Ritchie et al., 2016) and sequences (Ortega et al.”
Likelihood-free MCMC with Approximate Likelihood Ratios (2019, background): “Amortized inference [40, 41] expands on this idea by using generative models to capture inference mappings.”
Learning 3D Shape Completion from Laser Scan Data with Weak Supervision (2018, background): “Amortized Inference: The notion of amortized inference was introduced in [19] and exploited repeatedly in recent work [38, 40, 47].”
ProBO: Versatile Bayesian Optimization Using Any Probabilistic Programming Language (2019, background): “In amortized or compiled inference methods [39, 28, 24], inf trains or calls a pretrained model that maps observations to a posterior or proposal distribution, and post samples from this.”
Meta-Learning an Inference Algorithm for Probabilistic Programs (2021, background): “Amortised inference and inference compilation (Gershman & Goodman, 2014; Le et al., 2017; Paige & Wood, 2016; Stuhlmüller et al., 2013; Kingma & Welling, 2013; Mnih & Gregor, 2014; Rezende et al., 2014; Ritchie et al., 2016; Marino et al., 2018) are closely related to our approach in that they also ”
Monte Carlo Gradient Estimation in Machine Learning (2019, methodology): “Many of the approaches for automated inference in probabilistic programs, like variational inference, rely on the automated computation of Monte Carlo gradient estimates (Kucukelbir et al., 2017; Ritchie et al., 2016; Wingate et al., 2011).”
ProBO: a Framework for Using Probabilistic Programming in Bayesian Optimization (2019, methodology): “…those making use of Markov chain Monte Carlo (MCMC), most variational or exact inference methods (e.g. GPs and deep PPs (Tran et al., 2017)), and forward simulation methods (e.g. universal PPs (Le et al., 2016)), amortized inference (Ritchie et al., 2016), and ABC methods (Csill´ery et al., 2010)).”
MultiVerse: Causal Reasoning using Importance Sampling in Probabilistic Programming (2019, methodology): “…et al., 2015; Germain et al., 2015; Perov et al., 2015; Paige and Wood, 2016a; Perov, 2016; Le et al., 2016; Morris, 2001; Paige and Wood, 2016b; Ritchie et al., 2016; Le et al., 2017) and in particular importance sampling inference (Douglas et al., 2017; Walecki et al., 2018) can be used to…”
Simple, Distributed, and Accelerated Probabilistic Programming (2018, background): “First, with the exception of recent works [49, 36, 39, 42, 7, 34], most languages lack support for minibatch training and variational inference, and most lack critical systems features such as numerical stability, automatic differentiation, accelerator support, and vectorization.”
Likelihood-free MCMC with Amortized Approximate Ratio Estimators (2019, background): “Amortized variational inference (Gershman & Goodman, 2014; Ritchie et al., 2016) expands on this idea by using generative models to capture inference mappings.”