Sumon Kumar Bose
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Sumon Kumar Bose has a total of 38 co-authors. Here are the top co-authors by number of shared publications:
- Arindam Basu
City University of Hong Kong
12 shared publications - Mohendra Roy
Pandit Deendayal Energy University
3 shared publications Aakash Patil
Tork Motors, Pune 411026, India
2 shared publicationsBapi Kar
Delta-NTU Corporate Laboratory for Cyber-Physical Systems, School of EEE, Nanyang Technological University, Singapore 639798
2 shared publications- Bapi Kar
School of Electrical and Electronic Engineering Nanyang Technological University Singapore 639798
2 shared publications - Mohendra Roy
School of Electrical and Electronic Engineering Nanyang Technological University Singapore 639798
2 shared publications
Latest Publications
- 2 August 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Solid-State Circuits
- Vol. 58 (2) , 554-568
A 389 TOPS/W, Always ON Region Proposal Integrated Circuit Using In-Memory Computing in 65 nm CMOS
Neuromorphic vision sensors (NVSs) are key enablers of energy savings in Internet of Things (IoT)-based traffic monitoring and surveillance systems that exploit the temporal redundancy in video streams. However, for these scenarios, an object typically occupies a fraction of the full image frame leading to a significant spatial redundancy in the active image. Hence, there is a need for energy-efficient, dedicated hardware to detect the region of interests (RoI) to exploit spatial redundancy in the valid frames and reduce computations in the succeeding recognition modules. This article proposes a 9T-SRAM in-memory computing (IMC)-based region proposal (RP) network for event-based binary image (EBBI) frames from a NVS. The proposed 9T-SRAM cell enables a 1-D projection of objects on the horizontal and vertical axes of an image. An iterative and selective search (ISS) of the rising and falling edges of 1-D projection yields the coordinates of a bounding box encapsulating an object. To demonstrate the energy-saving and effectiveness of the algorithm, we fabricated the proposed architecture, RP integrated circuit (RPIC) in a 65 nm CMOS process. Tested with the video recordings from a Dynamic and Active-pixel Vision Sensor (DAVIS), the RPIC achieves a peak throughput of 1259 ft/s at 1 Meps event rate. Moreover, the proposed RP architecture achieves a high energy efficiency of 389 TOPS/W due to in-memory operation.
- 2 August 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Solid-State Circuits
- Vol. 58 (2) , 554-568
A 389 TOPS/W, Always ON Region Proposal Integrated Circuit Using In-Memory Computing in 65 nm CMOS
Neuromorphic vision sensors (NVSs) are key enablers of energy savings in Internet of Things (IoT)-based traffic monitoring and surveillance systems that exploit the temporal redundancy in video streams. However, for these scenarios, an object typically occupies a fraction of the full image frame leading to a significant spatial redundancy in the active image. Hence, there is a need for energy-efficient, dedicated hardware to detect the region of interests (RoI) to exploit spatial redundancy in the valid frames and reduce computations in the succeeding recognition modules. This article proposes a 9T-SRAM in-memory computing (IMC)-based region proposal (RP) network for event-based binary image (EBBI) frames from a NVS. The proposed 9T-SRAM cell enables a 1-D projection of objects on the horizontal and vertical axes of an image. An iterative and selective search (ISS) of the rising and falling edges of 1-D projection yields the coordinates of a bounding box encapsulating an object. To demonstrate the energy-saving and effectiveness of the algorithm, we fabricated the proposed architecture, RP integrated circuit (RPIC) in a 65 nm CMOS process. Tested with the video recordings from a Dynamic and Active-pixel Vision Sensor (DAVIS), the RPIC achieves a peak throughput of 1259 ft/s at 1 Meps event rate. Moreover, the proposed RP architecture achieves a high energy efficiency of 389 TOPS/W due to in-memory operation.
- 7 November 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-3
A 389TOPS/W, 1262fps at 1Meps Region Proposal Integrated Circuit for Neuromorphic Vision Sensors in 65nm CMOS
Neuromorphic vision sensors (NVS) [1] are key enablers in traffic monitoring and surveillance systems that exploit the temporal redundancy in video streams to get $\gt 2\mathrm{X}$ energy savings by blank frame detection (Fig. 1). Such concept of event driven processing has been used to reduce system energy for regular cameras as well [2]. However, an object typically occupies a fraction of the full image frame (Fig. 1) leading to a significant spatial redundancy in the image. Hence, an energy-efficient hardware is required to detect the region of interests (RoIs) in the valid frames to trigger an object recognition engine only for the RoIs. For a binary image, the region proposal can be performed by the connected component labeling (CCL) algorithm [2]. However, CCL scans the image in a raster fashion to calculate the ROIs leading to longer execution time and higher energy dissipation due to enormous data transfer. On the contrary, emerging in-memory [3], [4] and near-memory [5] computing approaches are a way to eliminate the data transfer cost and latency, promising further energy savings. In this paper, we propose 9T-SRAM based near and in-memory computing region proposal integrated circuit (RPIC) leveraging the $1 -\mathrm{D}$ projections of the objects on the vertical and horizontal axes. Further, we propose an iterative and selective search (ISS) algorithm to overcome overlapped projections among objects and provide an accurate number of objects and their exacts coordinates.- 7 November 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-3
A 389TOPS/W, 1262fps at 1Meps Region Proposal Integrated Circuit for Neuromorphic Vision Sensors in 65nm CMOS
Neuromorphic vision sensors (NVS) [1] are key enablers in traffic monitoring and surveillance systems that exploit the temporal redundancy in video streams to get $\gt 2\mathrm{X}$ energy savings by blank frame detection (Fig. 1). Such concept of event driven processing has been used to reduce system energy for regular cameras as well [2]. However, an object typically occupies a fraction of the full image frame (Fig. 1) leading to a significant spatial redundancy in the image. Hence, an energy-efficient hardware is required to detect the region of interests (RoIs) in the valid frames to trigger an object recognition engine only for the RoIs. For a binary image, the region proposal can be performed by the connected component labeling (CCL) algorithm [2]. However, CCL scans the image in a raster fashion to calculate the ROIs leading to longer execution time and higher energy dissipation due to enormous data transfer. On the contrary, emerging in-memory [3], [4] and near-memory [5] computing approaches are a way to eliminate the data transfer cost and latency, promising further energy savings. In this paper, we propose 9T-SRAM based near and in-memory computing region proposal integrated circuit (RPIC) leveraging the $1 -\mathrm{D}$ projections of the objects on the vertical and horizontal axes. Further, we propose an iterative and selective search (ISS) algorithm to overcome overlapped projections among objects and provide an accurate number of objects and their exacts coordinates.- 29 July 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Solid-State Circuits
- Vol. 57 (1) , 323-335
A 51.3-TOPS/W, 134.4-GOPS In-Memory Binary Image Filtering in 65-nm CMOS
Neuromorphic vision sensors (NVSs) can enable energy savings due to their event-driven that exploits the temporal redundancy in video streams from a stationary camera. However, noise-driven events lead to the false triggering of the object recognition processor. Image denoise operations require memory-intensive processing leading to a bottleneck in energy and latency. In this article, we present in-memory filtering (IMF), a 6T-SRAM in-memory computing (IMC)-based image denoising for event-based binary image (EBBI) frame from an NVS. We propose a non-overlap median filter (NOMF) for image denoising. An IMC framework enables hardware implementation of NOMF leveraging the inherent read disturb phenomenon of 6T-SRAM. To demonstrate the energy-saving and effectiveness of the algorithm, we fabricated the proposed architecture in a 65-nm CMOS process. Compared to fully digital implementation, IMF enables >70x energy savings and a >3x improvement of processing time when tested with the video recordings from a DAVIS sensor and achieves a peak throughput of 134.4 GOPS. Furthermore, the peak energy efficiencies of the NOMF are 51.3 TOPS/W, comparable with state-of-the-art in-memory processors. We also show that the accuracy of the images obtained by NOMF provides comparable accuracy in tracking and classification applications compared with images obtained by conventional median filtering.
- 29 July 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Solid-State Circuits
- Vol. 57 (1) , 323-335
A 51.3-TOPS/W, 134.4-GOPS In-Memory Binary Image Filtering in 65-nm CMOS
Neuromorphic vision sensors (NVSs) can enable energy savings due to their event-driven that exploits the temporal redundancy in video streams from a stationary camera. However, noise-driven events lead to the false triggering of the object recognition processor. Image denoise operations require memory-intensive processing leading to a bottleneck in energy and latency. In this article, we present in-memory filtering (IMF), a 6T-SRAM in-memory computing (IMC)-based image denoising for event-based binary image (EBBI) frame from an NVS. We propose a non-overlap median filter (NOMF) for image denoising. An IMC framework enables hardware implementation of NOMF leveraging the inherent read disturb phenomenon of 6T-SRAM. To demonstrate the energy-saving and effectiveness of the algorithm, we fabricated the proposed architecture in a 65-nm CMOS process. Compared to fully digital implementation, IMF enables >70x energy savings and a >3x improvement of processing time when tested with the video recordings from a DAVIS sensor and achieves a peak throughput of 134.4 GOPS. Furthermore, the peak energy efficiencies of the NOMF are 51.3 TOPS/W, comparable with state-of-the-art in-memory processors. We also show that the accuracy of the images obtained by NOMF provides comparable accuracy in tracking and classification applications compared with images obtained by conventional median filtering.
- 2 September 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Very Large Scale Integration (VLSI) Systems
- Vol. 28 (12) , 2518-2529
ADIC: Anomaly Detection Integrated Circuit in 65-nm CMOS Utilizing Approximate Computing
In this article, we present a low-power (LP) anomaly detection integrated circuit (ADIC) based on a one-class classifier (OCC) neural network. The ADIC achieves LP operation through a combination of: 1) careful choice of algorithm for online learning and 2) approximate computing techniques to lower average energy. In particular, online pseudoinverse update method (OPIUM) is used to train a randomized neural network for quick and resource-efficient learning. An additional 42% energy saving can be achieved when a lighter version of OPIUM method is used for training with the same number of data samples lead to no significant compromise on the quality of inference. Instead of a single classifier with large number of neurons, an ensemble of K base learner (BL) approach is chosen to reduce learning memory by a factor of K. This also enables approximate computing by dynamically varying the neural network size based on anomaly detection. Fabricated in 65-nm CMOS, the ADIC has K=7 BLs with 32 neurons in each BL and dissipates 11.87 and 3.35 pJ/OP during learning and inference, respectively, at Vdd=0.75,,V when all seven BLs are enabled. Furthermore, evaluated on the NASA bearing data set, approximately 80% of the chip can be shut down for 99% of the lifetime leading to an energy efficiency of 0.48 pJ/OP, an 18.5x reduction over full-precision computing running at Vdd=1.2 V throughout the lifetime.
- 2 September 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Very Large Scale Integration (VLSI) Systems
- Vol. 28 (12) , 2518-2529
ADIC: Anomaly Detection Integrated Circuit in 65-nm CMOS Utilizing Approximate Computing
In this article, we present a low-power (LP) anomaly detection integrated circuit (ADIC) based on a one-class classifier (OCC) neural network. The ADIC achieves LP operation through a combination of: 1) careful choice of algorithm for online learning and 2) approximate computing techniques to lower average energy. In particular, online pseudoinverse update method (OPIUM) is used to train a randomized neural network for quick and resource-efficient learning. An additional 42% energy saving can be achieved when a lighter version of OPIUM method is used for training with the same number of data samples lead to no significant compromise on the quality of inference. Instead of a single classifier with large number of neurons, an ensemble of K base learner (BL) approach is chosen to reduce learning memory by a factor of K. This also enables approximate computing by dynamically varying the neural network size based on anomaly detection. Fabricated in 65-nm CMOS, the ADIC has K=7 BLs with 32 neurons in each BL and dissipates 11.87 and 3.35 pJ/OP during learning and inference, respectively, at Vdd=0.75,,V when all seven BLs are enabled. Furthermore, evaluated on the NASA bearing data set, approximately 80% of the chip can be shut down for 99% of the lifetime leading to an energy efficiency of 0.48 pJ/OP, an 18.5x reduction over full-precision computing running at Vdd=1.2 V throughout the lifetime.
- Open Access
ADIC: Anomaly Detection Integrated Circuit in 65nm CMOS utilizing Approximate Computing
In this paper, we present a low-power anomaly detection integrated circuit (ADIC) based on a one-class classifier (OCC) neural network. The ADIC achieves low-power operation through a combination of (a) careful choice of algorithm for online learning and (b) approximate computing techniques to lower average energy. In particular, online pseudoinverse update method (OPIUM) is used to train a randomized neural network for quick and resource efficient learning. An additional 42% energy saving can be achieved when a lighter version of OPIUM method is used for training with the same number of data samples lead to no significant compromise on the quality of inference. Instead of a single classifier with large number of neurons, an ensemble of K base learner approach is chosen to reduce learning memory by a factor of K. This also enables approximate computing by dynamically varying the neural network size based on anomaly detection. Fabricated in 65nm CMOS, the ADIC has K = 7 Base Learners (BL) with 32 neurons in each BL and dissipates 11.87pJ/OP and 3.35pJ/OP during learning and inference respectively at Vdd = 0.75V when all 7 BLs are enabled. Further, evaluated on the NASA bearing dataset, approximately 80% of the chip can be shut down for 99% of the lifetime leading to an energy efficiency of 0.48pJ/OP, an 18.5 times reduction over full-precision computing running at Vdd = 1.2V throughout the lifetime.
Open AccessADIC: Anomaly Detection Integrated Circuit in 65nm CMOS utilizing Approximate Computing
In this paper, we present a low-power anomaly detection integrated circuit (ADIC) based on a one-class classifier (OCC) neural network. The ADIC achieves low-power operation through a combination of (a) careful choice of algorithm for online learning and (b) approximate computing techniques to lower average energy. In particular, online pseudoinverse update method (OPIUM) is used to train a randomized neural network for quick and resource efficient learning. An additional 42% energy saving can be achieved when a lighter version of OPIUM method is used for training with the same number of data samples lead to no significant compromise on the quality of inference. Instead of a single classifier with large number of neurons, an ensemble of K base learner approach is chosen to reduce learning memory by a factor of K. This also enables approximate computing by dynamically varying the neural network size based on anomaly detection. Fabricated in 65nm CMOS, the ADIC has K = 7 Base Learners (BL) with 32 neurons in each BL and dissipates 11.87pJ/OP and 3.35pJ/OP during learning and inference respectively at Vdd = 0.75V when all 7 BLs are enabled. Further, evaluated on the NASA bearing dataset, approximately 80% of the chip can be shut down for 99% of the lifetime leading to an energy efficiency of 0.48pJ/OP, an 18.5 times reduction over full-precision computing running at Vdd = 1.2V throughout the lifetime.
- Open Access
- 12 August 2020
- Published by Springer Nature in Nature Communications
- Vol. 11 (1) , 1-12
Self healable neuromorphic memtransistor elements for decentralized sensory signal processing in robotics
Sensory information processing in robot skins currently rely on a centralized approach where signal transduction (on the body) is separated from centralized computation and decision-making, requiring the transfer of large amounts of data from periphery to central processors, at the cost of wiring, latency, fault tolerance and robustness. We envision a decentralized approach where intelligence is embedded in the sensing nodes, using a unique neuromorphic methodology to extract relevant information in robotic skins. Here we specifically address pain perception and the association of nociception with tactile perception to trigger the escape reflex in a sensorized robotic arm. The proposed system comprises self-healable materials and memtransistors as enabling technologies for the implementation of neuromorphic nociceptors, spiking local associative learning and communication. Configuring memtransistors as gated-threshold and -memristive switches, the demonstrated system features in-memory edge computing with minimal hardware circuitry and wiring, and enhanced fault tolerance and robustness.
- 12 August 2020
- Published by Springer Nature in Nature Communications
- Vol. 11 (1) , 1-12
Open AccessSelf healable neuromorphic memtransistor elements for decentralized sensory signal processing in robotics
Sensory information processing in robot skins currently rely on a centralized approach where signal transduction (on the body) is separated from centralized computation and decision-making, requiring the transfer of large amounts of data from periphery to central processors, at the cost of wiring, latency, fault tolerance and robustness. We envision a decentralized approach where intelligence is embedded in the sensing nodes, using a unique neuromorphic methodology to extract relevant information in robotic skins. Here we specifically address pain perception and the association of nociception with tactile perception to trigger the escape reflex in a sensorized robotic arm. The proposed system comprises self-healable materials and memtransistors as enabling technologies for the implementation of neuromorphic nociceptors, spiking local associative learning and communication. Configuring memtransistors as gated-threshold and -memristive switches, the demonstrated system features in-memory edge computing with minimal hardware circuitry and wiring, and enhanced fault tolerance and robustness.
- Open Access
- 25 June 2020
- Published by Springer Nature in Nature Communications
- Vol. 11 (1) , 1-9
Optogenetics inspired transition metal dichalcogenide neuristors for in-memory deep recurrent neural networks
Shallow feed-forward networks are incapable of addressing complex tasks such as natural language processing that require learning of temporal signals. To address these requirements, we need deep neuromorphic architectures with recurrent connections such as deep recurrent neural networks. However, the training of such networks demand very high precision of weights, excellent conductance linearity and low write-noise- not satisfied by current memristive implementations. Inspired from optogenetics, here we report a neuromorphic computing platform comprised of photo-excitable neuristors capable of in-memory computations across 980 addressable states with a high signal-to-noise ratio of 77. The large linear dynamic range, low write noise and selective excitability allows high fidelity opto-electronic transfer of weights with a two-shot write scheme, while electrical in-memory inference provides energy efficiency. This method enables implementing a memristive deep recurrent neural network with twelve trainable layers with more than a million parameters to recognize spoken commands with >90% accuracy.
- 25 June 2020
- Published by Springer Nature in Nature Communications
- Vol. 11 (1) , 1-9
Open AccessOptogenetics inspired transition metal dichalcogenide neuristors for in-memory deep recurrent neural networks
Shallow feed-forward networks are incapable of addressing complex tasks such as natural language processing that require learning of temporal signals. To address these requirements, we need deep neuromorphic architectures with recurrent connections such as deep recurrent neural networks. However, the training of such networks demand very high precision of weights, excellent conductance linearity and low write-noise- not satisfied by current memristive implementations. Inspired from optogenetics, here we report a neuromorphic computing platform comprised of photo-excitable neuristors capable of in-memory computations across 980 addressable states with a high signal-to-noise ratio of 77. The large linear dynamic range, low write noise and selective excitability allows high fidelity opto-electronic transfer of weights with a two-shot write scheme, while electrical in-memory inference provides energy efficiency. This method enables implementing a memristive deep recurrent neural network with twelve trainable layers with more than a million parameters to recognize spoken commands with >90% accuracy.
- Open Access
A 75kb SRAM in 65nm CMOS for In-Memory Computing Based Neuromorphic Image Denoising
This paper presents an in-memory computing (IMC) architecture for image denoising. The proposed SRAM based in-memory processing framework works in tandem with approximate computing on a binary image generated from neuromorphic vision sensors. Implemented in TSMC 65nm process, the proposed architecture enables approximately 2000X energy savings (approximately 222X from IMC) compared to a digital implementation when tested with the video recordings from a DAVIS sensor and achieves a peak throughput of 1.25-1.66 frames/us.
Open AccessA 75kb SRAM in 65nm CMOS for In-Memory Computing Based Neuromorphic Image Denoising
This paper presents an in-memory computing (IMC) architecture for image denoising. The proposed SRAM based in-memory processing framework works in tandem with approximate computing on a binary image generated from neuromorphic vision sensors. Implemented in TSMC 65nm process, the proposed architecture enables approximately 2000X energy savings (approximately 222X from IMC) compared to a digital implementation when tested with the video recordings from a DAVIS sensor and achieves a peak throughput of 1.25-1.66 frames/us.
- 20 December 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Circuits and Systems I: Regular Papers
- Vol. 67 (3) , 913-926
ADEPOS: A Novel Approximate Computing Framework for Anomaly Detection Systems and its Implementation in 65-nm CMOS
To overcome the energy and bandwidth limitations of traditional IoT systems, ``edge computing'' or information extraction at the sensor node has become popular. However, now it is important to create very low energy information extraction or pattern recognition systems. In this paper, we present an approximate computing method to reduce the computation energy of a specific type of IoT system used for anomaly detection (e.g. in predictive maintenance, epileptic seizure detection, etc). Termed as Anomaly Detection Based Power Savings (ADEPOS), our proposed method uses low precision computing and low complexity neural networks at the beginning when it is easy to distinguish healthy data. However, on the detection of anomalies, the complexity of the network and computing precision are adaptively increased for accurate predictions. We show that ensemble approaches are well suited for adaptively changing network size. To validate our proposed scheme, a chip has been fabricated in UMC 65 nm process that includes an MSP430 microprocessor along with an on-chip switching mode DC-DC converter for dynamic voltage and frequency scaling. Using NASA bearing dataset for machine health monitoring, we show that using ADEPOS we can achieve 8.95X saving of energy along the lifetime without losing any detection accuracy. The energy savings are obtained by reducing the execution time of the neural network on the microprocessor.
- 20 December 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Circuits and Systems I: Regular Papers
- Vol. 67 (3) , 913-926
ADEPOS: A Novel Approximate Computing Framework for Anomaly Detection Systems and its Implementation in 65-nm CMOS
To overcome the energy and bandwidth limitations of traditional IoT systems, ``edge computing'' or information extraction at the sensor node has become popular. However, now it is important to create very low energy information extraction or pattern recognition systems. In this paper, we present an approximate computing method to reduce the computation energy of a specific type of IoT system used for anomaly detection (e.g. in predictive maintenance, epileptic seizure detection, etc). Termed as Anomaly Detection Based Power Savings (ADEPOS), our proposed method uses low precision computing and low complexity neural networks at the beginning when it is easy to distinguish healthy data. However, on the detection of anomalies, the complexity of the network and computing precision are adaptively increased for accurate predictions. We show that ensemble approaches are well suited for adaptively changing network size. To validate our proposed scheme, a chip has been fabricated in UMC 65 nm process that includes an MSP430 microprocessor along with an on-chip switching mode DC-DC converter for dynamic voltage and frequency scaling. Using NASA bearing dataset for machine health monitoring, we show that using ADEPOS we can achieve 8.95X saving of energy along the lifetime without losing any detection accuracy. The energy savings are obtained by reducing the execution time of the neural network on the microprocessor.
- Open Access
ADEPOS: A Novel Approximate Computing Framework for Anomaly Detection Systems and its Implementation in 65nm CMOS
To overcome the energy and bandwidth limitations of traditional IoT systems, edge computing or information extraction at the sensor node has become popular. However, now it is important to create very low energy information extraction or pattern recognition systems. In this paper, we present an approximate computing method to reduce the computation energy of a specific type of IoT system used for anomaly detection (e.g. in predictive maintenance, epileptic seizure detection, etc). Termed as Anomaly Detection Based Power Savings (ADEPOS), our proposed method uses low precision computing and low complexity neural networks at the beginning when it is easy to distinguish healthy data. However, on the detection of anomalies, the complexity of the network and computing precision are adaptively increased for accurate predictions. We show that ensemble approaches are well suited for adaptively changing network size. To validate our proposed scheme, a chip has been fabricated in UMC65nm process that includes an MSP430 microprocessor along with an on-chip switching mode DC-DC converter for dynamic voltage and frequency scaling. Using NASA bearing dataset for machine health monitoring, we show that using ADEPOS we can achieve 8.95X saving of energy along the lifetime without losing any detection accuracy. The energy savings are obtained by reducing the execution time of the neural network on the microprocessor.
Open AccessADEPOS: A Novel Approximate Computing Framework for Anomaly Detection Systems and its Implementation in 65nm CMOS
To overcome the energy and bandwidth limitations of traditional IoT systems, edge computing or information extraction at the sensor node has become popular. However, now it is important to create very low energy information extraction or pattern recognition systems. In this paper, we present an approximate computing method to reduce the computation energy of a specific type of IoT system used for anomaly detection (e.g. in predictive maintenance, epileptic seizure detection, etc). Termed as Anomaly Detection Based Power Savings (ADEPOS), our proposed method uses low precision computing and low complexity neural networks at the beginning when it is easy to distinguish healthy data. However, on the detection of anomalies, the complexity of the network and computing precision are adaptively increased for accurate predictions. We show that ensemble approaches are well suited for adaptively changing network size. To validate our proposed scheme, a chip has been fabricated in UMC65nm process that includes an MSP430 microprocessor along with an on-chip switching mode DC-DC converter for dynamic voltage and frequency scaling. Using NASA bearing dataset for machine health monitoring, we show that using ADEPOS we can achieve 8.95X saving of energy along the lifetime without losing any detection accuracy. The energy savings are obtained by reducing the execution time of the neural network on the microprocessor.