arXiv subjects

Barak A. Pearlmutter

Publications and source records attributed to Barak A. Pearlmutter.

At least 37 records · Page 2Linked to original sources

arXiv2019

Lock-Free Hopscotch Hashing

In this paper we present a lock-free version of Hopscotch Hashing. Hopscotch Hashing is an open addressing algorithm originally proposed by Herlihy, Shavit, and Tzafrir, which is known for fast performance and excellent cache locality. The algorithm allows users of the table to skip or jump over irrelevant entries, allowing quick search, insertion, and removal of entries. Unlike traditional linear probing, Hopscotch Hashing is capable of operating under a high load factor, as probe counts remain small. Our lock-free version improves on both speed, cache locality, and progress guarantees of the original, being a chimera of two concurrent hash tables. We compare our data structure to various other lock-free and blocking hashing algorithms and show that its performance is in many cases superior to existing strategies. The proposed lock-free version overcomes some of the drawbacks associated with the original blocking version, leading to a substantial boost in scalability while maintaining attractive features like physical deletion or probe-chain compression.

cs.DC↗

arXiv2019

Confusion of Tagged Perturbations in Forward Automatic Differentiation of Higher-Order Functions

Forward Automatic Differentiation (AD) is a technique for augmenting programs to compute derivatives. The essence of Forward AD is to attach perturbations to each number, and propagate these through the computation. When derivatives are nested, the distinct derivative calculations, and their associated perturbations, must be distinguished. This is typically accomplished by creating a unique tag for each derivative calculation, tagging the perturbations, and overloading the arithmetic operators. We exhibit a subtle bug, present in fielded implementations, in which perturbations are confused despite the tagging machinery. The essence of the bug is this: each invocation of a derivative creates a unique tag but a unique tag is needed for each derivative calculation. When taking derivatives of higher-order functions, these need not correspond! The derivative of a higher-order function $f$ that returns a function $g$ will be a function $f'$ that returns a function $\bar{g}$ that performs a derivative calculation. A single invocation of $f'$ will create a single fresh tag but that same tag will be used for each derivative calculation resulting from an invocation of $\bar{g}$. This situation arises when taking derivatives of curried functions. Two potential solutions are presented, and their serious deficiencies discussed. One requires eta expansion to delay the creation of fresh tags from the invocation of $f'$ to the invocation of $\bar{g}$, which can be difficult or even impossible in some circumstances. The other requires $f'$ to wrap $\bar{g}$ with tag renaming, which is difficult to implement without violating the desirable complexity properties of forward AD.

cs.SC↗

arXiv2018

Concurrent Robin Hood Hashing

In this paper we examine the issues involved in adding concurrency to the Robin Hood hash table algorithm. We present a non-blocking obstruction-free K-CAS Robin Hood algorithm which requires only a single word compare-and-swap primitive, thus making it highly portable. The implementation maintains the attractive properties of the original Robin Hood structure, such as a low expected probe length, capability to operate effectively under a high load factor and good cache locality, all of which are essential for high performance on modern computer architectures. We compare our data-structures to various other lock-free and concurrent algorithms, as well as a simple hardware transactional variant, and show that our implementation performs better across a number of contexts.

cs.DC↗

arXiv2018

Divide-and-Conquer Checkpointing for Arbitrary Programs with No User Annotation

Classical reverse-mode automatic differentiation (AD) imposes only a small constant-factor overhead in operation count over the original computation, but has storage requirements that grow, in the worst case, in proportion to the time consumed by the original computation. This storage blowup can be ameliorated by checkpointing, a process that reorders application of classical reverse-mode AD over an execution interval to tradeoff space \vs\ time. Application of checkpointing in a divide-and-conquer fashion to strategically chosen nested execution intervals can break classical reverse-mode AD into stages which can reduce the worst-case growth in storage from linear to sublinear. Doing this has been fully automated only for computations of particularly simple form, with checkpoints spanning execution intervals resulting from a limited set of program constructs. Here we show how the technique can be automated for arbitrary computations. The essential innovation is to apply the technique at the level of the language implementation itself, thus allowing checkpoints to span any execution interval.

cs.PL↗

arXiv2018

Automatic differentiation in machine learning: a survey

Derivatives, mostly in the form of gradients and Hessians, are ubiquitous in machine learning. Automatic differentiation (AD), also called algorithmic differentiation or simply "autodiff", is a family of techniques similar to but more general than backpropagation for efficiently and accurately evaluating derivatives of numeric functions expressed as computer programs. AD is a small but established field with applications in areas including computational fluid dynamics, atmospheric sciences, and engineering design optimization. Until very recently, the fields of machine learning and AD have largely been unaware of each other and, in some cases, have independently discovered each other's results. Despite its relevance, general-purpose AD has been missing from the machine learning toolbox, a situation slowly changing with its ongoing adoption under the names "dynamic computational graphs" and "differentiable programming". We survey the intersection of AD and machine learning, cover applications where AD has direct relevance, and address the main implementation techniques. By precisely defining the main differentiation techniques and their interrelationships, we aim to bring clarity to the usage of the terms "autodiff", "automatic differentiation", and "symbolic differentiation" as these are encountered more and more in machine learning settings.

cs.SC↗

arXiv2017

Nonnegative Factorization of a Data Matrix as a Motivational Example for Basic Linear Algebra

We present a motivating example for matrix multiplication based on factoring a data matrix. Traditionally, matrix multiplication is motivated by applications in physics: composing rigid transformations, scaling, sheering, etc. We present an engaging modern example which naturally motivates a variety of matrix manipulations, and a variety of different ways of viewing matrix multiplication. We exhibit a low-rank non-negative decomposition (NMF) of a "data matrix" whose entries are word frequencies across a corpus of documents. We then explore the meaning of the entries in the decomposition, find natural interpretations of intermediate quantities that arise in several different ways of writing the matrix product, and show the utility of various matrix operations. This example gives the students a glimpse of the power of an advanced linear algebraic technique used in modern data science.

math.HO↗

arXiv2016

Binomial Checkpointing for Arbitrary Programs with No User Annotation

Heretofore, automatic checkpointing at procedure-call boundaries, to reduce the space complexity of reverse mode, has been provided by systems like Tapenade. However, binomial checkpointing, or treeverse, has only been provided in Automatic Differentiation (AD) systems in special cases, e.g., through user-provided pragmas on DO loops in Tapenade, or as the nested taping mechanism in adol-c for time integration processes, which requires that user code be refactored. We present a framework for applying binomial checkpointing to arbitrary code with no special annotation or refactoring required. This is accomplished by applying binomial checkpointing directly to a program trace. This trace is produced by a general-purpose checkpointing mechanism that is orthogonal to AD.

cs.PL↗

arXiv2016

Efficient Implementation of a Higher-Order Language with Built-In AD

We show that Automatic Differentiation (AD) operators can be provided in a dynamic language without sacrificing numeric performance. To achieve this, general forward and reverse AD functions are added to a simple high-level dynamic language, and support for them is included in an aggressive optimizing compiler. Novel technical mechanisms are discussed, which have the ability to migrate the AD transformations from run-time to compile-time. The resulting system, although only a research prototype, exhibits startlingly good performance. In fact, despite the potential inefficiencies entailed by support of a functional-programming language and a first-class AD operator, performance is competitive with the fastest available preprocessor-based Fortran AD systems. On benchmarks involving nested use of the AD operators, it can even dramatically exceed their performance.

cs.PL↗

arXiv2016

DiffSharp: An AD Library for .NET Languages

DiffSharp is an algorithmic differentiation or automatic differentiation (AD) library for the .NET ecosystem, which is targeted by the C# and F# languages, among others. The library has been designed with machine learning applications in mind, allowing very succinct implementations of models and optimization routines. DiffSharp is implemented in F# and exposes forward and reverse AD operators as general nestable higher-order functions, usable by any .NET language. It provides high-performance linear algebra primitives---scalars, vectors, and matrices, with a generalization to tensors underway---that are fully supported by all the AD operators, and which use a BLAS/LAPACK backend via the highly optimized OpenBLAS library. DiffSharp currently uses operator overloading, but we are developing a transformation-based version of the library using F#'s "code quotation" metaprogramming facility. Work on a CUDA-based GPU backend is also underway.

cs.MS↗

arXiv2016

Evolving the Incremental λ Calculus into a Model of Forward Automatic Differentiation (AD)

Formal transformations somehow resembling the usual derivative are surprisingly common in computer science, with two notable examples being derivatives of regular expressions and derivatives of types. A newcomer to this list is the incremental $λ$-calculus, or ILC, a "theory of changes" that deploys a formal apparatus allowing the automatic generation of efficient update functions which perform incremental computation. The ILC is not only defined, but given a formal machine-understandable definition---accompanied by mechanically verifiable proofs of various properties, including in particular correctness of various sorts. Here, we show how the ILC can be mutated into propagating tangents, thus serving as a model of Forward Accumulation Mode Automatic Differentiation. This mutation is done in several steps. These steps can also be applied to the proofs, resulting in machine-checked proofs of the correctness of this model of forward AD.

cs.PL↗

arXiv2016

Tricks from Deep Learning

The deep learning community has devised a diverse set of methods to make gradient optimization, using large datasets, of large and highly complex models with deeply cascaded nonlinearities, practical. Taken as a whole, these methods constitute a breakthrough, allowing computational structures which are quite wide, very deep, and with an enormous number and variety of free parameters to be effectively optimized. The result now dominates much of practical machine learning, with applications in machine translation, computer vision, and speech recognition. Many of these methods, viewed through the lens of algorithmic differentiation (AD), can be seen as either addressing issues with the gradient itself, or finding ways of achieving increased efficiency using tricks that are AD-related, but not provided by current AD systems. The goal of this paper is to explain not just those methods of most relevance to AD, but also the technical constraints and mindset which led to their discovery. After explaining this context, we present a "laundry list" of methods developed by the deep learning community. Two of these are discussed in further mathematical detail: a way to dramatically reduce the size of the tape when performing reverse-mode AD on a (theoretically) time-reversible process like an ODE integrator; and a new mathematical insight that allows for the implementation of a stochastic Newton's method.

cs.LG↗

arXiv2015

DiffSharp: Automatic Differentiation Library

In this paper we introduce DiffSharp, an automatic differentiation (AD) library designed with machine learning in mind. AD is a family of techniques that evaluate derivatives at machine precision with only a small constant factor of overhead, by systematically applying the chain rule of calculus at the elementary operator level. DiffSharp aims to make an extensive array of AD techniques available, in convenient form, to the machine learning community. These including arbitrary nesting of forward/reverse AD operations, AD with linear algebra primitives, and a functional API that emphasizes the use of higher-order functions and composition. The library exposes this functionality through an API that provides gradients, Hessians, Jacobians, directional derivatives, and matrix-free Hessian- and Jacobian-vector products. Bearing the performance requirements of the latest machine learning techniques in mind, the underlying computations are run through a high-performance BLAS/LAPACK backend, using OpenBLAS by default. GPU support is currently being implemented.

cs.MS↗

arXiv2015

The Compositional Nature of Event Representations in the Human Brain

How does the human brain represent simple compositions of constituents: actors, verbs, objects, directions, and locations? Subjects viewed videos during neuroimaging (fMRI) sessions from which sentential descriptions of those videos were identified by decoding the brain representations based only on their fMRI activation patterns. Constituents (e.g., "fold" and "shirt") were independently decoded from a single presentation. Independent constituent classification was then compared to joint classification of aggregate concepts (e.g., "fold-shirt"); results were similar as measured by accuracy and correlation. The brain regions used for independent constituent classification are largely disjoint and largely cover those used for joint classification. This allows recovery of sentential descriptions of stimulus videos by composing the results of the independent constituent classifiers. Furthermore, classifiers trained on the words one set of subjects think of when watching a video can recognise sentences a different subject thinks of when watching a different video.

q-bio.NC↗

arXiv2014

An Analysis of Publication Venues for Automatic Differentiation Research

We present the results of our analysis of publication venues for papers on automatic differentiation (AD), covering academic journals and conference proceedings. Our data are collected from the AD publications database maintained by the autodiff.org community website. The database is purpose-built for the AD field and is expanding via submissions by AD researchers. Therefore, it provides a relatively noise-free list of publications relating to the field. However, it does include noise in the form of variant spellings of journal and conference names. We handle this by manually correcting and merging these variants under the official names of corresponding venues. We also share the raw data we get after these corrections.

cs.DL↗

arXiv2014

Automatic Differentiation of Algorithms for Machine Learning

Automatic differentiation---the mechanical transformation of numeric computer programs to calculate derivatives efficiently and accurately---dates to the origin of the computer age. Reverse mode automatic differentiation both antedates and generalizes the method of backwards propagation of errors used in machine learning. Despite this, practitioners in a variety of fields, including machine learning, have been little influenced by automatic differentiation, and make scant use of available tools. Here we review the technique of automatic differentiation, describe its two main modes, and explain how it can benefit machine learning practitioners. To reach the widest possible audience our treatment assumes only elementary differential calculus, and does not assume any knowledge of linear algebra.

cs.LG↗

arXiv2013

The Compositional Nature of Verb and Argument Representations in the Human Brain

How does the human brain represent simple compositions of objects, actors,and actions? We had subjects view action sequence videos during neuroimaging (fMRI) sessions and identified lexical descriptions of those videos by decoding (SVM) the brain representations based only on their fMRI activation patterns. As a precursor to this result, we had demonstrated that we could reliably and with high probability decode action labels corresponding to one of six action videos (dig, walk, etc.), again while subjects viewed the action sequence during scanning (fMRI). This result was replicated at two different brain imaging sites with common protocols but different subjects, showing common brain areas, including areas known for episodic memory (PHG, MTL, high level visual pathways, etc.,i.e. the 'what' and 'where' systems, and TPJ, i.e. 'theory of mind'). Given these results, we were also able to successfully show a key aspect of language compositionality based on simultaneous decoding of object class and actor identity. Finally, combining these novel steps in 'brain reading' allowed us to accurately estimate brain representations supporting compositional decoding of a complex event composed of an actor, a verb, a direction, and an object.

q-bio.NC↗

arXiv2013

Block Coordinate Descent for Sparse NMF

Nonnegative matrix factorization (NMF) has become a ubiquitous tool for data analysis. An important variant is the sparse NMF problem which arises when we explicitly require the learnt features to be sparse. A natural measure of sparsity is the L$_0$ norm, however its optimization is NP-hard. Mixed norms, such as L$_1$/L$_2$ measure, have been shown to model sparsity robustly, based on intuitive attributes that such measures need to satisfy. This is in contrast to computationally cheaper alternatives such as the plain L$_1$ norm. However, present algorithms designed for optimizing the mixed norm L$_1$/L$_2$ are slow and other formulations for sparse NMF have been proposed such as those based on L$_1$ and L$_0$ norms. Our proposed algorithm allows us to solve the mixed norm sparsity constraints while not sacrificing computation time. We present experimental evidence on real-world datasets that shows our new algorithm performs an order of magnitude faster compared to the current state-of-the-art solvers optimizing the mixed norm and is suitable for large-scale datasets.

cs.LG↗

arXiv2012

AD in Fortran, Part 1: Design

We propose extensions to Fortran which integrate forward and reverse Automatic Differentiation (AD) directly into the programming model. Irrespective of implementation technology, embedding AD constructs directly into the language extends the reach and convenience of AD while allowing abstraction of concepts of interest to scientific-computing practice, such as root finding, optimization, and finding equilibria of continuous games. Multiple different subprograms for these tasks can share common interfaces, regardless of whether and how they use AD internally. A programmer can maximize a function F by calling a library maximizer, XSTAR=ARGMAX(F,X0), which internally constructs derivatives of F by AD, without having to learn how to use any particular AD tool. We illustrate the utility of these extensions by example: programs become much more concise and closer to traditional mathematical notation. A companion paper describes how these extensions can be implemented by a program that generates input to existing Fortran-based AD tools.

cs.PL↗