Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Incremental Learning of Affordances using Markov Logic Networks

George Potter, Gertjan Burghouts, Joris Sijs

TL;DR

The MLN Cumulative Learning Algorithm (MLN-CLA) is introduced and it is shown that MLN-CLA is effective for accumulative learning and zero-shot affordance inference, outperforming strong baselines.

Abstract

Affordances enable robots to have a semantic understanding of their surroundings. This allows them to have more acting flexibility when completing a given task. Capturing object affordances in a machine learning model is a difficult task, because of their dependence on contextual information. Markov Logic Networks (MLN) combine probabilistic reasoning with logic that is able to capture such context. Mobile robots operate in partially known environments wherein unseen object affordances can be observed. This new information must be incorporated into the existing knowledge, without having to retrain the MLN from scratch. We introduce the MLN Cumulative Learning Algorithm (MLN-CLA). MLN-CLA learns new relations in various knowledge domains by retaining knowledge and only updating the changed knowledge, for which the MLN is retrained. We show that MLN-CLA is effective for accumulative learning and zero-shot affordance inference, outperforming strong baselines.

Incremental Learning of Affordances using Markov Logic Networks

TL;DR

The MLN Cumulative Learning Algorithm (MLN-CLA) is introduced and it is shown that MLN-CLA is effective for accumulative learning and zero-shot affordance inference, outperforming strong baselines.

Abstract

Affordances enable robots to have a semantic understanding of their surroundings. This allows them to have more acting flexibility when completing a given task. Capturing object affordances in a machine learning model is a difficult task, because of their dependence on contextual information. Markov Logic Networks (MLN) combine probabilistic reasoning with logic that is able to capture such context. Mobile robots operate in partially known environments wherein unseen object affordances can be observed. This new information must be incorporated into the existing knowledge, without having to retrain the MLN from scratch. We introduce the MLN Cumulative Learning Algorithm (MLN-CLA). MLN-CLA learns new relations in various knowledge domains by retaining knowledge and only updating the changed knowledge, for which the MLN is retrained. We show that MLN-CLA is effective for accumulative learning and zero-shot affordance inference, outperforming strong baselines.

Paper Structure

This paper contains 12 sections, 4 equations, 4 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Method
  4. Updating Knowledge Lists
  5. Updating Knowledge Categories
  6. Learning process
  7. Cumulative Learning Algorithm
  8. Experiments
  9. Learning new constants
  10. Learning new formulas
  11. Knowledge updating strategies
  12. Conclusions

Figures (4)

  • Figure 1: In batch learning, two MLN-database pairs are combined into one before training. In MLN-CLA, each MLN-database pair is trained separately before merging.
  • Figure 2: MLN-CLA: We build our knowledge categorisation and knowledge updating strategies on top of the same concepts as MCLA cui2022mcla, outlined in purple. With the blue outlined box we indicate the overlap with OSL huynh2011osl.
  • Figure 3: Learning of new constants and new formulas.
  • Figure 4: Marginal probability predictions of the three knowledge updating strategies over the steps in the constants learning experiment.

Theorems & Definitions (4)

  • Definition 1: Knowledge List
  • Example 1
  • Definition 2: Knowledge Category
  • Example 2