Conditional Deep Learning for energy-efficient and enhanced pattern recognition

By Priyadarshini Panda, Abhronil Sengupta, Kaushik Roy. 2015

208 citations recorded by the source corpus.

Deep learning neural networks have emerged as one of the most powerful classification tools for vision related applications. However, the computational and energy requirements associated with such deep nets can be quite high, and hence their energy-efficient implementation is of great interest. Although traditionally the entire network is utilized for the recognition of all inputs, we observe that the classification difficulty varies widely across inputs in real-world datasets; only a small fraction of inputs require the full computational effort of a network, while a large majority can be classified correctly with very low effort. In this paper, we propose Conditional Deep Learning (CDL) where the convolutional layer features are used to identify the variability in the difficulty of input instances and conditionally activate the deeper layers of the network. We achieve this by cascading a linear network of output neurons for each convolutional layer and monitoring the output of the linear network to decide whether classification can be terminated at the current stage or not. The proposed methodology thus enables the network to dynamically adjust the computational effort depending upon the difficulty of the input data while maintaining competitive classification accuracy. We evaluate our approach on the MNIST dataset. Our experiments demonstrate that our proposed CDL yields 1.91× reduction in average number of operations per input, which translates to 1.84× improvement in energy. In addition, our results show an improvement in classification accuracy from 97.5% to 98.9% as compared to the original network.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Model Architectures, Efficiency & Compression

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Vision and Pattern Recognition

In Model Architectures

In Efficiency & Compression

Citations in the corpus

Why Should We Add Early Exits to Neural Networks? (2020, neutral): “[ 72, 55, 12, 32 ] have considered a variation of LT, in which the backbone network is trained separately from each of the auxiliary classiﬁer.”
BranchyNet: Fast inference via early exiting from deep neural networks (2016, background): “[18] propose Conditional Deep Learning (CDL) by iteratively adding linear classifiers to each convolutional layer, starting with the first layer, and monitoring the output to decide whether a sample can be exited early.”
Designing Adaptive Neural Networks for Energy-Constrained Image Classification (2018, background): “ and complexity. This insight of dynamically trading off accuracy with energy-efficiency can be found in several existing approaches from different groups and under different names (e.g., conditional [18], scalable [32], etc.). Theearlyeffortstoenableenergyefficiencywerebasedon“earlyexit” conditions”
Energy-Efficient and Improved Image Recognition with Conditional Deep Learning (2017, methodology): “The key idea of appending additional output layers has been proposed earlier in our previous work [Panda et al. 2016].”
S2DNAS: Transforming Static CNN Model for Dynamic Inference via Neural Architecture Search (2019, methodology): “To mitigate these problems, researchers propose early exiting the “easy” input samples at inference time [31, 40, 17, 1].”
Dynamic Convolutions: Exploiting Spatial Sparsity for Faster Inference (2019, background): “Another approach is to use early-exit branches in the network [7, 35, 45].”
PredictiveNet: An energy-efficient convolutional neural network via zero prediction (2017, background): “[10] proposed Conditional Deep Learning (CDL) by adding a linear network to each convolutional layer and monitoring the output to decide whether classification can be terminated at the current stage.”
Hardware-Software Co-design Approach for Deep Learning Inference (2019, background): “In [10], Conditional Deep Learning (CDL) algorithm is presented which have added linear classifiers after each of the convolutional layers and monitors the prediction confidence of the linear layers to determine whether the classification can be terminated or not.”