aiqbench analysis
This paper is primarily about machine learning methods: adaptive sparse evolutionary training and warm-start compression of small convolutional classifiers. Its energy-monitoring setting is the application, not the subject. The reported gains are modest (macro F1 up to 92.40% with 17% fewer stored weights), and the authors explicitly caution against claims of statistical equivalence or performance on unseen houses. Readers interested in edge-deployable models should note that the compression gains come with additional training cost and rest on synthetic aggregate data.
Evolutionary Deep Nets for Non-Intrusive Load Monitoring
arXiv:2303.03538v2 Announce Type: replace Abstract: Compact appliance classifiers can support local electricity-load monitoring where memory and computation are constrained. We evaluate adaptive sparse evolutionary training and warm-start compression using synthetic aggregate windows derived from household appliance recordings. A completed convolutional network study trains 120 models over four targets and three seeds with validation-only model selection. Two analyses examine compression with approximately preserved binary accuracy and joint improvement in positive-class F1 and sparsity. Relative to a newly trained fixed sparse evolutionary baseline, the selected compressed policy changes accuracy from 91.53%, while increasing macro F1 to 92.40%. Stored weights decrease by 17.36% and active first-layer connections by 23.93%. Refinement requires additional training. Three seeds and unverified source-window independence prevent claims of statistical equivalence or unseen-house performance.
Abstract:Compact appliance classifiers can support local electricity-load monitoring where memory and computation are constrained. We evaluate adaptive sparse evolutionary training and warm-start compression using synthetic aggregate windows derived from household appliance recordings. A completed convolutional network study trains 120 models over four targets and three seeds with validation-only model selection. Two analyses examine compression with approximately preserved binary accuracy and joint improvement in positive-class F1 and sparsity. Relative to a newly trained fixed sparse evolutionary baseline, the selected compressed policy changes accuracy from 91.53%, while increasing macro F1 to 92.40%. Stored weights decrease by 17.36% and active first-layer connections by 23.93%. Refinement requires additional training. Three seeds and unverified source-window independence prevent claims of statistical equivalence or unseen-house performance.
Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Neural and Evolutionary Computing (cs.NE); Signal Processing (eess.SP); Systems and Control (eess.SY)
arXiv:2303.03538 [cs.LG]
(or arXiv:2303.03538v2 [cs.LG] for this version)
https://doi.org/10.13140/RG.2.2.25983.28324 DOI(s) linking to related resources
Full text under the publisher's open licence · 151 words