# Think Again Networks and the Delta Loss

## TL;DR

This paper introduces Think Again Networks (ThinkNet), an abstraction applicable to state-dependent functions like recurrent neural networks, aiming to enhance their flexibility and performance.

## Contribution

The paper proposes ThinkNet, a novel abstraction that can be integrated into various state-dependent models to improve their adaptability.

## Key findings

- ThinkNet effectively models state-dependent functions.
- Application of ThinkNet improves neural network performance.
- The abstraction is versatile across different neural architectures.

## Abstract

This short paper introduces an abstraction called Think Again Networks (ThinkNet) which can be applied to any state-dependent function (such as a recurrent neural network).