Deleting object selective units in a fully-connected layer of deep convolutional networks improves classification performance
Neurons in the primate visual cortices show a wide range of stimulus selectivity. Some neurons respond to only a small fraction of stimulus images, whereas others respond to many stimulus images in a non-selective manner. It is unclear how stimulus selective and non-selective neurons contribute to visual object recognition. Herein, we examined the relationship between stimulus selectivity and the effect of deletion of units on task performance using fully a connected layer of two types of deep convolutional neural networks (DCNNs). Deleting a stimulus selective unit caused slight improvements of task performance, whereas deleting stimulus non-selective units caused a significant decrease in task performance. However, these findings do not imply that stimulus selective units have no use for the task. Indeed, better performance was obtained when the networks consisted of both stimulus selective and non-selective units.
Code (0)
등록된 구현이 없습니다.
Tasks
Object RecognitionSimilar Papers 제목 키워드 기반
Causal importance of orientation selectivity for generalization in image recognition
Although both our brain and deep neural networks (DNNs) can perform high-level sensory-perception tasks such as image or speech recognition, the inner mechanism of these hierarchical information-processing systems is poo…
image-classificationImage ClassificationObject Recognitionspeech-recognitionOn-the-fly Network Pruning for Object Detection
Object detection with deep neural networks is often performed by passing a few thousand candidate bounding boxes through a deep neural network for each image. These bounding boxes are highly correlated since they origina…
Network PruningObjectobject-detectionObject DetectionA note on 'A fully parallel 3D thinning algorithm and its applications'
A 3D thinning algorithm erodes a 3D binary image layer by layer to extract the skeletons. This paper presents a correction to Ma and Sonka's thinning algorithm, A fully parallel 3D thinning algorithm and its applications…
Are there any 'object detectors' in the hidden layers of CNNs trained to identify objects or scenes?
Various methods of measuring unit selectivity have been developed with the aim of better understanding how neural networks work. But the different measures provide divergent estimates of selectivity, and this has led to …
General Classificationimage-classificationImage ClassificationObjectLow Latency CMOS Hardware Acceleration for Fully Connected Layers in Deep Neural Networks
We present a novel low latency CMOS hardware accelerator for fully connected (FC) layers in deep neural networks (DNNs). The FC accelerator, FC-ACCL, is based on 128 8x8 or 16x16 processing elements (PEs) for matrix-vect…