Log In Sign Up

Routing Networks and the Challenges of Modular and Compositional Computation

by   Clemens Rosenbaum, et al.

Compositionality is a key strategy for addressing combinatorial complexity and the curse of dimensionality. Recent work has shown that compositional solutions can be learned and offer substantial gains across a variety of domains, including multi-task learning, language modeling, visual question answering, machine comprehension, and others. However, such models present unique challenges during training when both the module parameters and their composition must be learned jointly. In this paper, we identify several of these issues and analyze their underlying causes. Our discussion focuses on routing networks, a general approach to this problem, and examines empirically the interplay of these challenges and a variety of design decisions. In particular, we consider the effect of how the algorithm decides on module composition, how the algorithm updates the modules, and if the algorithm uses regularization.


Neural Module Networks

Visual question answering is fundamentally compositional in nature---a q...

Question Guided Modular Routing Networks for Visual Question Answering

Visual Question Answering (VQA) faces two major challenges: how to bette...

Semantic-aware Modular Capsule Routing for Visual Question Answering

Visual Question Answering (VQA) is fundamentally compositional in nature...

Modular Lifelong Reinforcement Learning via Neural Composition

Humans commonly solve complex problems by decomposing them into easier s...

Visual Question Reasoning on General Dependency Tree

The collaborative reasoning for understanding each image-question pair i...

Compositionality as Lexical Symmetry

Standard deep network models lack the inductive biases needed to general...

Modular Networks: Learning to Decompose Neural Computation

Scaling model capacity has been vital in the success of deep learning. F...