CoChat

FilterNet: A Many-to-Many Deep Learning Architecture for Time Series Classification

Open access

Preprint — published as 10.3390/s20092498

The DOI matches its Crossref record; 1 of the 1 references it cites were found in a registry or an index.

Abstract

We present and benchmark FilterNet, a flexible deep learning architecture for time series classification tasks, such as activity recognition via multichannel sensor data. It adapts popular CNN and CNN-LSTM motifs which have excelled in activity recognition benchmarks, implementing them in a many-to-many architecture to markedly improve frame-by-frame accuracy, event segmentation accuracy, model size, and computational efficiency. We propose several model variants, evaluate them alongside other published models using the Opportunity benchmark dataset, demonstrate the effect of model ensembling and of altering key parameters, and quantify the quality of the models’ segmentation of discrete events. We also offer recommendations for use and suggest potential model extensions. FilterNet advances the state of the art in all measured accuracy and speed metrics on the benchmarked dataset, and it can be extensively customized for other applications.

Overview

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.20944/preprints202002.0318.v2

Cites

1 of 1 references verified0 citation errors · 0 not checkable

Statistics

Figures

Figure 1 In a typical many-to-one approach (left) an input is first divided into fixed-length overlapping windows, then a model processes each window individually, generating a class prediction for each one, and finally the predictions are concatenated into an output time series. In a many-to-many approach (right), the entire output time series is generated with a single model evaluation.
Figure 1 In a typical many-to-one approach (left) an input is first divided into fixed-length overlapping windows, then a model processes each window individually, generating a class prediction for each one, and finally the predictions are concatenated into an output time series. In a many-to-many approach (right), the entire output time series is generated with a single model evaluation.CC BY · PubMed Central
Figure 2 FilterNet architecture. (a) Each FilterNet model is composed primarily of one or more stacks of FilterNet layer modules (FLMs), which are parameterized and constructed as shown (see text for elaboration). (b) In the prototypical FilterNet component architecture, FLMs are grouped into components that can be parameterized and combined to implement time series classifiers of varying speed and complexity, tuned to the problem at hand.
Figure 2 FilterNet architecture. (a) Each FilterNet model is composed primarily of one or more stacks of FilterNet layer modules (FLMs), which are parameterized and constructed as shown (see text for elaboration). (b) In the prototypical FilterNet component architecture, FLMs are grouped into components that can be parameterized and combined to implement time series classifiers of varying speed and complexity, tuned to the problem at hand.CC BY · PubMed Central
Figure 3 Representative training history for a ms-C/L model. While the validation loss oscillated and had near-global minima at epochs 27, 35, and 41, the custom stopping metric (see text) adjusted more predictably to a minimum at epoch 36. Training was stopped at epoch 47 and the model from epoch 36 was restored and used for subsequent inference.
Figure 3 Representative training history for a ms-C/L model. While the validation loss oscillated and had near-global minima at epochs 27, 35, and 41, the custom stopping metric (see text) adjusted more predictably to a minimum at epoch 36. Training was stopped at epoch 47 and the model from epoch 36 was restored and used for subsequent inference.CC BY · PubMed Central
Figure 4 Heatmap demonstrating differences between model outputs. (a) The ground-truth labels for 150 s during the first run in the standard Opportunity test set; (b–e) alongside the corresponding predictions for the 17 non-null behavior classes for various FilterNet architectures. Panes are annotated with weighted F1 scores (F1w and the event-based F1e) calculated over the plotted region.
Figure 4 Heatmap demonstrating differences between model outputs. (a) The ground-truth labels for 150 s during the first run in the standard Opportunity test set; (b–e) alongside the corresponding predictions for the 17 non-null behavior classes for various FilterNet architectures. Panes are annotated with weighted F1 scores (F1w and the event-based F1e) calculated over the plotted region.CC BY · PubMed Central
Figure 5 Event-based metrics. Performance metrics for several classifiers, including event-based precision Pe, recall Re, and F1 score F1e, alongside event summary diagrams, each for a single representative run.
Figure 5 Event-based metrics. Performance metrics for several classifiers, including event-based precision Pe, recall Re, and F1 score F1e, alongside event summary diagrams, each for a single representative run.CC BY · PubMed Central
Figure 6 Performance of n-fold ensembled ms-C/L models. Both sample-based (a) and event-based (b) weighted F1 metrics improve substantially with the number of ensembled models, plateauing between 3–5 folds, while inference rate (pane c) decreases. (mean ± sd, n = 10).
Figure 6 Performance of n-fold ensembled ms-C/L models. Both sample-based (a) and event-based (b) weighted F1 metrics improve substantially with the number of ensembled models, plateauing between 3–5 folds, while inference rate (pane c) decreases. (mean ± sd, n = 10).CC BY · PubMed Central
Figure 7 Effects of inference window length. For all models, accuracy metrics (top panes) improve as inference window length increases, especially when models incorporate multi-scale or LSTM architectural features. For LSTM-containing models, the inference rate (bottom panes) falls off for long windows because their calculation across the time dimension cannot be fully parallelized. (n = 5 each).
Figure 7 Effects of inference window length. For all models, accuracy metrics (top panes) improve as inference window length increases, especially when models incorporate multi-scale or LSTM architectural features. For LSTM-containing models, the inference rate (bottom panes) falls off for long windows because their calculation across the time dimension cannot be fully parallelized. (n = 5 each).CC BY · PubMed Central
Figure 8 Performance of different Opportunity sensor subsets (mean of n = 5 runs) according to (a) the sample-by-sample F1w,nn and (b) the event-based F1 F1e. Using larger sensor subsets, including gyroscopes (G), accelerometers (A), and the magnetic (Mag) components of the inertial measurement units, as well as all 113 standard sensors channels (All), tended to improve performance metrics. The best models (e.g., 4× ms-C/L, ms-C/L, ms-CNN, and p-C/L) maintain relatively high performance even with fewer sensor channels.
Figure 8 Performance of different Opportunity sensor subsets (mean of n = 5 runs) according to (a) the sample-by-sample F1w,nn and (b) the event-based F1 F1e. Using larger sensor subsets, including gyroscopes (G), accelerometers (A), and the magnetic (Mag) components of the inertial measurement units, as well as all 113 standard sensors channels (All), tended to improve performance metrics. The best models (e.g., 4× ms-C/L, ms-C/L, ms-CNN, and p-C/L) maintain relatively high performance even with fewer sensor channels.CC BY · PubMed Central

Verification

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Citations

Metadata

Authors
Chambers, R. D., & Yoder, N. C.
Corresponding author
Chambers, Robert D.
Venue
Sensors
Publisher
MDPI AG
Published
2020-04-15
Volume
20
Issue
9
Pages
2498-2498
ISSN
1424-8220 (ISSN-L)
OpenAlex
W3008141589
Cited by
17
References
1
Citations by year
2025: 1 · 2024: 2 · 2023: 6 · 2022: 3 · 2021: 3 · 2020: 2
FWCI
1.36
Type
journal article
Language
en
Open access
Locations
  • Sensors (publishedVersion, cc-by)
  • Preprints.org (acceptedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
Topics
  • Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Context-Aware Activity Recognition Systems
  • Physical Sciences › Computer Science › Signal Processing › Time Series Analysis and Forecasting
  • Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Neural Network Applications
Keywords
time series classification, activity recognition, convolutional neural networks, long short-term memory, event segmentation, model ensembling, computational efficiency, Opportunity dataset

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.