Machine Learning on Mainstream Microcontrollers
Open accessThe DOI matches its Crossref record; 27 of the 28 references it cites were found in a registry or an index.
Abstract
This paper presents the Edge Learning Machine (ELM), a machine learning framework for edge devices, which manages the training phase on a desktop computer and performs inferences on microcontrollers. The framework implements, in a platform-independent C language, three supervised machine learning algorithms (Support Vector Machine (SVM) with a linear kernel, k-Nearest Neighbors (K-NN), and Decision Tree (DT)), and exploits STM X-Cube-AI to implement Artificial Neural Networks (ANNs) on STM32 Nucleo boards. We investigated the performance of these algorithms on six embedded boards and six datasets (four classifications and two regression). Our analysis—which aims to plug a gap in the literature—shows that the target platforms allow us to achieve the same performance score as a desktop machine, with a similar time latency. ANN performs better than the other algorithms in most cases, with no difference among the target devices. We observed that increasing the depth of an NN improves performance, up to a saturation level. k-NN performs similarly to ANN and, in one case, even better, but requires all the training sets to be kept in the inference phase, posing a significant memory demand, which can be afforded only by high-end edge devices. DT performance has a larger variance across datasets. In general, several factors impact performance in different ways across datasets. This highlights the importance of a framework like ELM, which is able to train and compare different algorithms. To support the developer community, ELM is released on an open-source basis.
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.
Integrity signals
Registry
Cites
Statistics
Figures





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.
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
The source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“As a rough summary result, the best performance was that of ANN in all datasets except Sonar dataset. In contrast the SVM achieved the best results on the Heart and EnviroCar dataset. The performance of DT was on Virus and Peugeot 14. However, KNN achieved the best result in the Sonar dataset only.”
From TinyML: Enabling of Inference Deep Learning Models on Ultra-Low-Power IoT Edge Devices for AI Applications (2022) · cites this paper as [5]
Experimental — accuracy not yet measuredThe cited source says what the claim attributes to it.
In the citing paper
“Four algorithms were implemented on six devices from ARM Cortex-M microcontrollers released by STM-32, namely (F091RC, F303RE, F401RE, F746ZG, H743ZI2, and L452RE).”
From TinyML: Enabling of Inference Deep Learning Models on Ultra-Low-Power IoT Edge Devices for AI Applications (2022) · cites this paper as [5]
The cited source says what the claim attributes to it.
In the citing paper
“The authors [5] developed an Edge Learning Machine (ELM) framework to execute ML inference in edge devices such as Microcontroller.”
From TinyML: Enabling of Inference Deep Learning Models on Ultra-Low-Power IoT Edge Devices for AI Applications (2022) · cites this paper as [5]
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
Cites
Adam: A Method for Stochastic Optimization
Kingma, D. P., & Ba, J. (2014) · arXiv · doi:10.48550/arxiv.1412.6980 · Cited by 98,946
Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift
Ioffe, S., & Szegedy, C. (2015) · arXiv · doi:10.48550/arxiv.1502.03167 · Cited by 27,818
Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift
Ioffe, S. (2024) · arXiv (Cornell University) · doi:10.57702/o9raffed · Cited by 14,214
Internet of Things (IoT): A vision, architectural elements, and future directions
Gubbi, J., Buyya, R., Marusic, S., & Palaniswami, M. (2013) · Future Generation Computer Systems · doi:10.1016/j.future.2013.01.010 · Cited by 12,293 · cited 1×
“, [2]) to the edge has advantages in terms of response latency, bandwidth occupancy, energy consumption, security and expected privacy (e.”
BackgroundEdge computing: Vision and challenges
Shi, W., Cao, J., Zhang, Q., Li, Y., & Xu, L. (2016) · IEEE Internet of Things Journal · doi:10.1109/jiot.2016.2579198 · Cited by 8,733
Understanding Machine Learning: From Theory to Algorithms
Shalev-Shwartz, S., & Ben-David, S. (2014) · doi:10.1017/cbo9781107298019 · Cited by 3,115
(2007) · The MIT Press eBooks · doi:10.7551/mitpress/7496.001.0001 · Cited by 536
Computation Offloading Toward Edge Computing
Lin, L., Liao, X., Jin, H., & Li, P. (2019) · Proceedings of the IEEE · doi:10.1109/jproc.2019.2922285 · Cited by 456 · cited 1×
“The trend of moving computation towards the edge is becoming ever more relevant, leading to performance improvements and the development of new field data processing applications [1].”
BackgroundNearest-Neighbor Methods in Learning and Vision: Theory and Practice
(2006) · The MIT Press eBooks · doi:10.7551/mitpress/4908.001.0001 · Cited by 440
Generative Adversarial Networks for face generation: A survey
Kammoun, A., Slama, R., Tabia, H., Ouni, T., & Abid, M. (2022) · ACM Computing Surveys · doi:10.1145/1122445.1122456 · Cited by 311
Edge Machine Learning: Enabling Smart Internet of Things Applications
Yazici, M., Basurra, S., & Gaber, M. (2018) · Big Data and Cognitive Computing · doi:10.3390/bdcc2030026 · Cited by 105
Compiling KB-sized machine learning models to tiny IoT devices
Gopinath, S., Ghanathe, N., Seshadri, V., & Sharma, R. (2019) · Proceedings of the 40th ACM SIGPLAN Conference on Programming Language Design and Implementation · doi:10.1145/3314221.3314597 · Cited by 92 · cited 1×
“A major achievement concerns the translation of floating-point ML models into fixed-point code [23], which is, however, not the case in state-of-the-art mainstream microcontrollers.”
BackgroundFastGRNN: A Fast, Accurate, Stable and Tiny Kilobyte Sized Gated Recurrent Neural Network
Kusupati, A., Singh, M., Bhatia, K., Kumar, A., Jain, P., & Varma, M. (2019) · arXiv (Cornell University) · doi:10.48550/arxiv.1901.02358 · Cited by 91 · cited 1×
“Their work also concerns recurrent neural networks [22].”
BackgroundStatistical fingerprint‐based intrusion detection system (SF‐IDS)
Boero, L., Cello, M., Marchese, M., Mariconti, E., Naqash, T., & Zappatore, S. (2016) · International Journal of Communication Systems · doi:10.1002/dac.3225 · Cited by 75
Deep Learning: Edge-Cloud Data Analytics for IoT
Ghosh, A. M., & Grolinger, K. (2019) · 2019 IEEE Canadian Conference of Electrical and Computer Engineering (CCECE) · doi:10.1109/ccece.2019.8861806 · Cited by 62 · cited 1×
“[25] used autoencoders at the edge layer that are capable of dimensionality reduction to reduce the required processing time and storage space.”
MethodCerutti, G., Prasad, R., & Farella, E. (2019) · IEEE International Conference on Acoustics Speech and Signal Processing · doi:10.1109/icassp.2019.8682998 · Cited by 48 · cited 1×
“[19] implemented a convolutional neural network on STM Nucleo-L476RG for people detection using CMSIS-NN, which is an optimized library that allows for the deployment of NNs on Cortex-M microcontrollers.”
MethodSmart Tactile Sensing Systems Based on Embedded CNN Implementations
Alameh, M., Abbass, Y., Ibrahim, A., & Valle, M. (2020) · Micromachines · doi:10.3390/mi11010103 · Cited by 47 · cited 1×
“[30] created a smart tactile sensing system by implementing a convolutional neural network on various hardware platforms like Raspberry Pi 4, NVidia Jetson TX2, and Movidius NCS2 for tactile data decoding.”
MethodDeepEmote: Towards multi-layer neural networks in a low power wearable multi-sensors bracelet
Magno, M., Pritz, M., Mayer, P., & Benini, L. (2017) · 2017 7th IEEE International Workshop on Advances in Sensors and Interfaces (IWASI) · doi:10.1109/iwasi.2017.7974208 · Cited by 44
Are Existing Knowledge Transfer Techniques Effective for Deep Learning with Edge Devices?
Sharma, R., Biookaghazadeh, S., Li, B., & Zhao, M. (2018) · 2018 IEEE International Conference on Edge Computing (EDGE) · doi:10.1109/edge.2018.00013 · Cited by 38 · cited 1×
“[31] investigated the application of KT to edge devices, achieving good results by transferring knowledge from both the intermediate layers and the last layer of the teacher (original model) to a shallower student (target).”
BackgroundAnalyzing Machine Learning on Mainstream Microcontrollers
Falbo, V., Apicella, T., Aurioso, D., Danese, L., Bellotti, F., Berta, R., & Gloria, A. D. (2020) · Lecture notes in electrical engineering · doi:10.1007/978-3-030-37277-4_12 · Cited by 30
Self-citationA Simple Method for Solving the SVM Regularization Path for Semidefinite Kernels
Sentelle, C. G., Anagnostopoulos, G. C., & Georgiopoulos, M. (2015) · IEEE Transactions on Neural Networks and Learning Systems · doi:10.1109/tnnls.2015.2427333 · Cited by 29 · cited 1×
“In SVM, C is a key regularization parameter, that controls the tradeoff between errors of the SVM on training data and margin maximization [13,57].”
A Fuzzy Logic Module to Estimate a Driver’s Fuel Consumption for Reality-Enhanced Serious Games
Massoud, R., Poslad, S., Bellotti, F., Berta, R., Mehran, K., & Gloria, A. D. (2018) · International Journal of Serious Games · doi:10.17083/ijsg.v5i4.266 · Cited by 27 · cited 4×
“Heart 80% Tanh [300, 200, 100, 50] 30% Virus 99% ReLU [100, 100, 100] None Sonar 87% ReLU [300, 200, 100, 50] 30% Peugeot_Target 14 99% ReLU [100, 100, 100] mle Peugeot_Target 15 98% Tanh [300, 200, 100, 50] mle EnviroCar 0.”
BackgroundInfluentialSelf-citationThe Implementation of CNN-Based Object Detector on ARM Embedded Platforms
Zhang, Y., Bi, S., Dong, M., & Liu, Y. (2018) · IEEE Intl Conf on Dependable, Autonomic and Secure Computing, Intl Conf on Pervasive Intelligence and Computing, Intl Conf on Cloud and Big Data Computing, Intl Conf on Cyber Science and Technology Congress (DASC/PiCom/CBDCom/CyberSciTech) · doi:10.1109/dasc/picom/datacom/cyberscitec.2018.00074 · Cited by 19 · cited 1×
“[16] presented an object detector, namely MobileNet-Single Shot Detector (SSD), which was trained using a deep convolutional neural network with the popular Caffe framework.”
MethodMagno, M., Cavigelli, L., Mayer, P., von Hagen, F., & Benini, L. (2019) · IEEE World Forum on Internet of Things (WF-IoT) · doi:10.1109/wf-iot.2019.8767290 · Cited by 11 · cited 1×
“[27] presented an open-source toolkit, namely FANNCortexM.”
Practical Microcontroller Engineering with ARM® Technology
(2015) · Wiley eBooks · doi:10.1002/9781119058397 · Cited by 8 · cited 1×
“While the literature is increasingly reporting on novel or adapted embedded machine learning algorithms, architectures and applications, there is a lack of quantitative analyses about the performance of common ML algorithms on state-of-the-art mainstream edge devices, such as ARM microcontrollers [10].”
BackgroundDeveloping a Machine Learning Library for Microcontrollers
Parodi, A. F., Bellotti, F., Berta, R., & Gloria, A. D. (2019) · Lecture notes in electrical engineering · doi:10.1007/978-3-030-11973-7_36 · Cited by 5 · cited 1×
“a sonar system that analyses materials, distinguishing between rocks and metallic material [46,47].”
BackgroundSelf-citationImplementation details to reduce the latency of an SDN Statistical Fingerprint-Based IDS
Fausto, A., & Marchese, M. (2019) · International Symposium on Advanced Electrical and Communication Technologies (ISAECT) · doi:10.1109/isaect47714.2019.9069714 · Cited by 5
(2019) · Lecture Notes in Computer Science · doi:10.1007/978-3-030-34350-7 · Cited by 5
Research works printed but not linked: 3
- Suda N. Loh D. Machine Learning on ARM Cortex-M Microcontrollers Arm Ltd. Cambridge, UK 2019
- Goodfellow I. Bengio Y. Courville A. Deep Learning The MIT Press Cambridge, UK 2016
- Vapnik V.N. Lerner A.Y. Recognition of Patterns with help of Generalized Portraits Avtomat. Telemekh. 1963 24 774 780
Cited by
TinyML for Ultra-Low Power AI and Large Scale IoT Deployments: A Systematic Review
Schizas, N., Karras, A., Karras, C., & Sioutas, S. (2022) · Future Internet · doi:10.3390/fi14120363 · Cited by 215
Alajlan, N. N., & Ibrahim, D. M. (2022) · Micromachines · doi:10.3390/mi13060851 · Cited by 224
Quantization and Deployment of Deep Neural Networks on Microcontrollers
Novac, P.-E., Boukli Hacene, G., Pegatoquet, A., Miramond, B., & Gripon, V. (2021) · Sensors · doi:10.3390/s21092984 · Cited by 210
Artificial Intelligence in the IoT Era: A Review of Edge AI Hardware and Software
Sipola, T., Alatalo, J., Kokkonen, T., & Rantonen, M. (2022) · 2022 31st Conference of Open Innovations Association (FRUCT) · doi:10.23919/fruct54823.2022.9770931 · Cited by 100
“The weka-porter project supports only WEKA [177] decision tree conversions, and the Micro-LM project supports only models trained with the Desk-LM module [174], although the Desk-LM module in turn depends on and uses scikit-learn library.”
BackgroundMethodOn-Device Training of Machine Learning Models on Microcontrollers with Federated Learning
Giménez, N. L., Grau, M. M., Centelles, R. P., & Freitag, F. (2022) · Electronics · doi:10.3390/electronics11040573 · Cited by 77
Metadata
- Authors
- Sakr, F., Bellotti, F., Berta, R., & De Gloria, A.
- Affiliations
- University of Genoa (IT)
- Department of Electrical, Electronic and Telecommunication Engineering (DITEN)-University of Genoa, Via Opera Pia 11a, 16145 Genova, Italy
- Corresponding author
- Bellotti, Francesco
- Venue
- Sensors
- Publisher
- MDPI AG
- Published
- 2020-05-05
- Volume
- 20
- Issue
- 9
- Pages
- 2638
- ISSN
- 1424-8220 (ISSN-L)
- PMC
- PMC7249132
- OpenAlex
- W3021575660
- Cited by
- 109
- References
- 27
- Citations by year
- 2026: 10 · 2025: 14 · 2024: 25 · 2023: 16 · 2022: 21 · 2021: 20 · 2020: 2
- FWCI
- 6.93
- Type
- journal article
- Language
- en
- Open access
- gold
- dx.doi.org
- Licence
- Locations
- Sensors (publishedVersion, cc-by)
- PubMed Central (publishedVersion, cc-by)
- PubMed (publishedVersion)
- DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
- CINECA IRIS Institutial Research Information System (University of Genoa) (submittedVersion, other-oa)
- Topics
- Physical Sciences › Computer Science › Computer Networks and Communications › IoT and Edge/Fog Computing
- Physical Sciences › Engineering › Control and Systems Engineering › Embedded Systems and FPGA Design
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Neural Network Applications
- Keywords
- edge machine learning, microcontroller, embedded machine learning, artificial neural networks, support vector machine, k-nearest neighbors, decision tree, supervised classification, regression analysis, inference latency, open source software framework
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.