Neural Random Access Machines

By Karol Kurach, Marcin Andrychowicz, Ilya Sutskever. 2015

160 citations recorded by the source corpus.

Abstract: In this paper, we propose and investigate a new neural network architecture called Neural Random Access Machine. It can manipulate and dereference pointers to an external variable-size random-access memory. The model is trained from pure input-output examples using backpropagation. We evaluate the new model on a number of simple algorithmic tasks whose solutions require pointer manipulation and dereferencing. Our results show that the proposed model can learn to solve algorithmic tasks of such type and is capable of operating on simple data structures like linked-lists and binary trees. For easier tasks, the learned solutions generalize to sequences of arbitrary length. Moreover, memory access during inference can be done in a constant time under some assumptions.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Model Architectures

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

In Computer Vision and Pattern Recognition

In Model Architectures

Citations in the corpus

Show Your Work: Scratchpads for Intermediate Computation with Language Models (2021, neutral): “The related problem of neural algorithm induction has attracted considerable interest (Graves et al., 2014; Kurach et al., 2016; Kaiser & Sutskever, 2016; Graves et al., 2016; Reed & de Freitas, 2016; Veliˇckovi´c et al., 2020a;b).”
Program Synthesis with Large Language Models (2021, neutral): “…are deep network architectures that aim to learn algorithms from input-output examples, by structuring the network in a way that corresponds to mathematical models of computation like Turing machines [Graves et al., 2014, Kaiser and Sutskever, 2016, Kurach et al., 2016, Graves et al., 2016].”
Neural Execution Engines: Learning to Execute Subroutines (2020, background): “Recently, researchers have turned their attention towards imbuing neural networks with the capability to perform algorithmic reasoning , thereby allowing them to go beyond pattern recognition and logically solve more complex problems [9, 12, 10, 14].”
Making Neural Programming Architectures Generalize via Recursion (2017, background): “Numerous models have been proposed for learning programs; to name a few, this includes the Differentiable Neural Computer (Graves et al., 2016), Neural Turing Machine (Graves et al., 2014), Neural GPU (Kaiser & Sutskever, 2015), Neural Programmer (Neelakantan et al., 2015), Pointer Network (Vinyals ”
Least-to-Most Prompting Enables Complex Reasoning in Large Language Models (2022, background): “Notable examples of such tasks are: (1) Symbolic manipulation (Grefenstette et al., 2015; Joulin & Mikolov, 2015; Kaiser & Sutskever, 2016; Kurach et al., 2016; Graves et al., 2016; Dong et al., 2019), which usually involves manipulating lists of ∗Correspondence to: Denny Zhou <dennyzhou@google.”
Neural Programmer-Interpreters (2015, neutral): “(Zaremba & Sutskever, 2015; Zaremba et al., 2015; Kaiser & Sutskever, 2015; Kurach et al., 2015; Neelakantan et al., 2015).”
Incorporating Copying Mechanism in Sequence-to-Sequence Learning (2016, methodology): “…tends to cover consecutive words, including UNK s. Unlike the explicit de-sign for hybrid addressing in Neural Turing Machine (Graves et al., 2014; Kurach et al., 2015), C OPY N ET is more subtle: it provides the architecture that can facilitate some particular location-based addressing and lets…”
RobustFill: Neural Program Learning under Noisy I/O (2017, methodology): “…Riedel et al., 2016), and (2) neural program induction , where network learns to generate the output directly using a latent program representation (Graves et al., 2014; 2016; Kurach et al., 2016; Kaiser & Sutskever, 2015; Joulin & Mikolov, 2015; Reed & de Fre-itas, 2016; Neelakantan et al., 2016).”
WikiReading: A Novel Large-scale Language Understanding Task over Wikipedia (2016, methodology): “3 Instead of the linearization method of Sukhbaatar et al. (2015), we applied an entropy regularizer for the softmax attention as described in Kurach et al. (2015). by this attention (eqn.”
Inferring and Executing Programs for Visual Reasoning (2017, background): “Such models can take the form of a feedforward scoring function over operators in a domain-specific language that can be used to guide program search [4], or of a recurrent network that decodes a vectorial program representation into the actual program [22, 26, 34, 46, 47, 48].”
DeepCoder: Learning to Write Programs (2016, background): “Recently, there has been much interest in program-like neural network models (Graves et al., 2014; Weston et al., 2015; Kurach et al., 2015; Joulin & Mikolov, 2015; Grefenstette et al., 2015; Sukhbaatar et al., 2015; Neelakantan et al., 2016; Kaiser & Sutskever, 2016; Reed & de Freitas, 2016;…”
Learning to Optimize (2016, background): “Subsequent work proposes variants of this model that use different primitive memory access operations [14], more expressive operations [16, 28] or other non-differentiable operations [30, 29].”
On the Binding Problem in Artificial Neural Networks (2020, background): “In contrast, Reed and de Freitas (2015) and Kurach et al. (2016) take a step towards modern computer architectures by, respectively, incorporating a call-stack with an explicit compositional structure or a mechanism for manipulating and dereferencing pointers to a differentiable memory tape.”
Relational inductive biases, deep learning, and graph networks (2018, methodology): “Other methods have attempted to capture different types of structure by mimicking key hardware and software components in computers and how they transfer information between each other, such as persistent slotted storage, registers, memory I/O controllers, stacks, and queues (e.g. Dyer et al., 2015;”
Programmatically Interpretable Reinforcement Learning (2018, background): “These architectures typically augment neural networks with differentiable computational substrates such as memory (Neural Turing Machines (Graves et al., 2014)), modules (Neural RAM (Kurach et al., 2015)) or data-structures such as stacks (Joulin & Mikolov, 2015), and formulate the program learning ”
Adding Gradient Noise Improves Learning for Very Deep Networks (2015, methodology): “Because of this complexity, training the NRAM architecture can be unstable, especially when the number of steps and operations is large.”
Neuro-Symbolic Program Synthesis (2016, methodology): “While some recently proposed methods (Kurach et al., 2015; Gaunt et al., 2016; Riedel et al., 2016; Bunel et al., 2016) do learn interpretable programs, they still need to learn a separate neural network model for each individual task.”
Online and Linear-Time Attention by Enforcing Monotonic Alignments (2017, background): “Combining all of the above, we present our differentiable approach to training the monotonic alignment decoder in algorithm 2 (appendix A).”
Programming with a Differentiable Forth Interpreter (2016, background): “Many of these models were based on differentiable versions of abstract data structures (Joulin & Mikolov, 2015; Grefenstette et al., 2015; Kurach et al., 2016), and a few abstract machines, such as the NTM (Graves et al.”
Automatically Composing Representation Transformations as a Means for Generalization (2018, methodology): “Compositional approaches (as opposed to memory-augmented [6, 33– 35, 45, 51, 80] or monolithic [46, 89] approaches for learning programs) to the challenge of discovering reusable primitive transformations and their means of composition generally fall into two categories.”