Fully Integrated Analog Machine Learning Classifier Using Custom Activation Function for Low Resolution Image Classification

Chandrasekaran, Sanjeev Tannirkulam; Jayaraj, Akshay; Karnam, Vinay Elkoori Ghantala; Banerjee, Imon; Sanyal, Arindam

doi:10.1109/tcsi.2020.3047331

Cited by 17 publications

(15 citation statements)

References 26 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Instead of doing error correction at circuit level, which increases area and power cost, we use an error-aware ANN design methodology that takes into account the difference between ideal activation functions and the actual activation functions implemented using analog circuits to produce high accuracy AI predictions. In our prior study that focused on an image classification task 36 , 37 we developed an error-aware hardware-software co-design interface in which SPICE simulations are used to characterize a unit activation function, and the characterization data is imported into Matlab for training the ANN (see Fig. 7 e).…”

Section: Methodsmentioning

confidence: 99%

Fusion of fully integrated analog machine learning classifier with electronic medical records for real-time prediction of sepsis onset

Sadasivuni

Saha

Banerjee

et al. 2022

Sci Rep

Self Cite

View full text Add to dashboard Cite

The objective of this work is to develop a fusion artificial intelligence (AI) model that combines patient electronic medical record (EMR) and physiological sensor data to accurately predict early risk of sepsis. The fusion AI model has two components—an on-chip AI model that continuously analyzes patient electrocardiogram (ECG) data and a cloud AI model that combines EMR and prediction scores from on-chip AI model to predict fusion sepsis onset score. The on-chip AI model is designed using analog circuits for sepsis prediction with high energy efficiency for integration with resource constrained wearable device. Combination of EMR and sensor physiological data improves prediction performance compared to EMR or physiological data alone, and the late fusion model has an accuracy of 93% in predicting sepsis 4 h before onset. The key differentiation of this work over existing sepsis prediction literature is the use of single modality patient vital (ECG) and simple demographic information, instead of comprehensive laboratory test results and multiple vital signs. Such simple configuration and high accuracy makes our solution favorable for real-time, at-home use for self-monitoring.

show abstract

Section: Methodsmentioning

confidence: 99%

Fusion of fully integrated analog machine learning classifier with electronic medical records for real-time prediction of sepsis onset

Sadasivuni

Saha

Banerjee

et al. 2022

Sci Rep

Self Cite

View full text Add to dashboard Cite

show abstract

“…The custom analog activation functions resemble their ideal, mathematical counterparts, but are not exactly the same. To ensure good matching between software ANN model and IC measurements, we use a hardware-software co-design methodology in which amplifier transfer curves, and their derivatives, are used to train the ANN model iteratively 38 . Stochastic gradient descent is used to optimize the ANN model by minimizing the loss function at each epoch.…”

Section: Methodsmentioning

confidence: 99%

In-sensor neural network for high energy efficiency analog-to-information conversion

Sadasivuni

Bhanushali

Banerjee

et al. 2022

Sci Rep

View full text Add to dashboard Cite

This work presents an on-chip analog-to-information conversion technique that utilizes analog hyper-dimensional computing based on reservoir-computing paradigm to process electrocardiograph (ECG) signals locally in-sensor and reduce radio frequency transmission by more than three orders-of-magnitude. Instead of transmitting the naturally sparse ECG signal or extracted features, the on-chip analog-to-information converter analyzes the ECG signal through a nonlinear reservoir kernel followed by an artificial neural network, and transmits the prediction results. The proposed technique is demonstrated for detection of sepsis onset and achieves state-of-the-art accuracy and energy efficiency while reducing sensor power by $$159\times $$ 159 × with test-chips prototyped in 65 nm CMOS.

show abstract

“…Sigmoid function, Tanh function, and Relu function are widely used activation functions [25][26][27][28].…”

Section: Cnnmentioning

confidence: 99%

“…Sigmoid function, Tanh function, and Relu function are widely used activation functions [ 25 – 28 ]. The image of activation functions is exhibited in Figure 5 .…”

Section: Investment Risk and DL Model Analysis Of Multinational Enter...mentioning

confidence: 99%

Application of Deep Learning Model in the Avoidance of Investment Risk of Multinational Enterprises

Yan

Feng

2022

Computational Intelligence and Neuroscience

View full text Add to dashboard Cite

With the continuous improvement and development of the socialist market economic system, China's economic development has full momentum, but the domestic market is no longer sufficient to meet the needs of enterprise development. China has always focused on peaceful diplomacy, and the world market has a strong demand for Chinese products. This work aims to improve the accuracy of exchange rate forecasting. The risk factors that may be encountered in the investment process of multinational enterprises can be effectively avoided. Combining the advantages of Long Short-Term Memory (LSTM) and Convolutional Neural Network (CNN), the LSTM-CNN (Long Short-Term Memory-Convolutional Neural Network) model is proposed to predict the volatility trend of stocks. Firstly, the investment risk of multinational enterprises is analyzed, and, secondly, the principles of the used CNN and LSTM are expounded. Finally, the performance of the proposed model is verified by setting experiments. The experimental results demonstrate that when predicting the 10 selected stocks, the proposed LSTM-CNN model has the highest accuracy in predicting the volatility of stocks, with an average accuracy of 60.1%, while the average accuracy of the rest of the models is all below 60%. It can be found that the stock category does not have a great impact on the prediction accuracy of the model. The average prediction accuracy of the CNN model is 0.578, which is lower than that of the Convolutional Neural Network-Relevance model, and the prediction accuracy of the LSTM model is 0.592, which is better than that of the Long Short-Term Memory-Relevance model. The designed model can be used to predict the stock market to guide investors to make effective investments and reduce investment risks based on relevant cases. The research makes a certain contribution to improving the company's income and stabilizing the national economic development.

show abstract

Fully Integrated Analog Machine Learning Classifier Using Custom Activation Function for Low Resolution Image Classification

Cited by 17 publications

References 26 publications

Fusion of fully integrated analog machine learning classifier with electronic medical records for real-time prediction of sepsis onset

Fusion of fully integrated analog machine learning classifier with electronic medical records for real-time prediction of sepsis onset

In-sensor neural network for high energy efficiency analog-to-information conversion

Application of Deep Learning Model in the Avoidance of Investment Risk of Multinational Enterprises

Contact Info

Product

Resources

About