MouldIt: Data Efficiency of Multi-Task Temporal Representation Learning in Injection Moulding
MouldIt: Data Efficiency of Multi-Task Temporal Representation Learning in Injection Moulding
This research paper introduces MouldIt, a model using temporal convolution and attention mechanisms for injection moulding process analysis. The authors synthesize a dataset based on an Erlanger Tr"ager part cycle and evaluate the model's ability to reconstruct signals and classify parts against classical methods like k-nearest neighbors and random forests. Results show MouldIt outperforms baselines in both single-modal and multi-modal settings, achieving specific sorting losses of 6.0% +/- 0.7% and 1.0% +/- 0.3% respectively.
The study evaluates a temporal convolution and attention-based variational autoencoder named MouldIt for injection moulding, demonstrating improved data efficiency over classical methods starting from 2000 to 4000 cycles depending on modality.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
Reported numbers
cycles
4,000
View original evidence
odal and multi-modal settings. We compare it with an accept-all baseline and the classical methods k-nearest neighbors, logistic regression, random forests, and principal component analysis, which we adapt to support both tasks. MouldIt outperforms classical methods across both tasks from 4000 cycles in the single-modal setting and from 2000 cycles in the multi-modal setting. It achieves part sorting losses of 6.0 $\
Open source S4cycles
2,000
View original evidence
odal and multi-modal settings. We compare it with an accept-all baseline and the classical methods k-nearest neighbors, logistic regression, random forests, and principal component analysis, which we adapt to support both tasks. MouldIt outperforms classical methods across both tasks from 4000 cycles in the single-modal setting and from 2000 cycles in the multi-modal setting. It achieves part sorting losses of 6.0 $\
Open source S4
- paper: not_reported
- code: not_reported
- weights: not_reported
- dataset: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
ess-informed dataset based on a real reference cycle of an Erlanger Tr\"ager part. We then introduce MouldIt, a temporal convolution and attention based variational autoencoder that reconstructs in-mould signals, while classifying parts as acceptable or as one of our synthetic failure modes. We evaluate MouldIt on two tasks, known-class classification and unknown-class detection, across nine dataset sizes in single-m
Open source S3
odal and multi-modal settings. We compare it with an accept-all baseline and the classical methods k-nearest neighbors, logistic regression, random forests, and principal component analysis, which we adapt to support both tasks. MouldIt outperforms classical methods across both tasks from 4000 cycles in the single-modal setting and from 2000 cycles in the multi-modal setting. It achieves part sorting losses of 6.0 $\
Open source S4
pm$ 0.7% and 1.0 $\pm$ 0.3% in the single-modal and multi-modal settings respectively, ranking first across all baselines and a combination of principal component analysis and random forests.
Open source S5
Source：arXiv Robotics — research abstracts · arxiv.org