Comparison of semantic vectors with reduced precision using the cosine similarity measure

Karwatowski, Michał; Wielgosz, Maciej; Pietroń, Marcin; Staruchowicz, Mateusz; Wiatr, K.

doi:10.1109/intellisys.2017.8324236

“…[8]- [10] shown that Cosine similarity measurement (CSM) has a good representation in measuring distances between vectors. Even [8] showing cosine similarity measure has an accuracy of up to 0.99 in measuring semantic vectors that have been reduced to 8 bits. Moreover [9] also shows that CSM faster than the linear scan and approximation methods.…”

Section: Introductionmentioning

confidence: 99%

The Improved Artificial Neural Network Based on Cosine Similarity in Facial Emotion Recognition

Kirana¹,

Wibawanto²,

Hidayah³

et al. 2019

EECSI

View full text Add to dashboard Cite

In this study, we present the improved artificial neural network based on cosine similarity in facial emotion recognition. We apply a shifting window that employs a neural network for two concurrent processes consisting of face detection and emotional recognition. To prevent the slow and futile computations, non-face areas need to be filtered from neurons on each network layer, thus we propose the improved artificial neural network based on cosine similarity. Cosine similarity is employed to bypass the process of non-face areas in the neural network. The accuracy of the proposed method reaches 0.84, while the accuracy of the original neural network method reaches 0.74. It can be concluded that our methods work accurately. It can be concluded that our method works accurately. The proposed method is superior to the state-of-theart algorithms

show abstract

“…The measurement of the similarity between vectors is required in the filter of the non-face area. [8]- [10] shown that Cosine similarity measurement (CSM) has a good representation in measuring distances between vectors. Even [8] showing cosine similarity measure has an accuracy of up to 0.99 in measuring semantic vectors that have been reduced to 8 bits.…”

Section: Introductionmentioning

confidence: 99%

“…The CSM measures the cosine of the angle between two non-zero vectors. In other words, CSM measures the similarity between two non-zero vectors [8]- [10]. Therefore, we present the improved artificial neural network based on cosine similarity in facial emotion recognition.…”

Section: Introductionmentioning

confidence: 99%

The Improved Artificial Neural Network Based on Cosine Similarity in Facial Emotion Recognition

Kirana¹,

Wibawanto²,

Hidayah³

et al. 2019

EECSI

0

View full text Add to dashboard Cite

In this study, we present the improved artificial neural network based on cosine similarity in facial emotion recognition. We apply a shifting window that employs a neural network for two concurrent processes consisting of face detection and emotional recognition. To prevent the slow and futile computations, non-face areas need to be filtered from neurons on each network layer, thus we propose the improved artificial neural network based on cosine similarity. Cosine similarity is employed to bypass the process of non-face areas in the neural network. The accuracy of the proposed method reaches 0.84, while the accuracy of the original neural network method reaches 0.74. It can be concluded that our methods work accurately. It can be concluded that our method works accurately. The proposed method is superior to the state-of-theart algorithms

show abstract

“…However, creating efficient design architecture and its implementation are not trivial and generate interesting research task. As authors of this paper already began work on the dedicated hardware platform and presented their initial results in [20], we will not cover this topic. Still much effort needs to be put into FPGA implementation in order to utilize its potential in NLP tasks.…”

Section: Introductionmentioning

confidence: 99%

Improving Text Classification with Vectors of Reduced Precision

Wróbel

¹

,

Wielgosz

²

,

Pietroń

³

et al. 2018

Proceedings of the 10th International Conference on Agents and Artificial Intelligence

Self Cite

5

0

View full text Add to dashboard Cite

This paper presents the analysis of the impact of a floating-point number precision reduction on the quality of text classification. The precision reduction of the vectors representing the data (e.g. TF-IDF representation in our case) allows for a decrease of computing time and memory footprint on dedicated hardware platforms. The impact of precision reduction on the classification quality was performed on 5 corpora, using 4 different classifiers. Also, dimensionality reduction was taken into account. Results indicate that the precision reduction improves classification accuracy for most cases (up to 25% of error reduction). In general, the reduction from 64 to 4 bits gives the best scores and ensures that the results will not be worse than with the full floating-point representation.Precision reduction approach may not have significant per-

show abstract

Comparison of semantic vectors with reduced precision using the cosine similarity measure

Cited by 4 publications

References 12 publications

The Improved Artificial Neural Network Based on Cosine Similarity in Facial Emotion Recognition

The Improved Artificial Neural Network Based on Cosine Similarity in Facial Emotion Recognition

The Improved Artificial Neural Network Based on Cosine Similarity in Facial Emotion Recognition

Improving Text Classification with Vectors of Reduced Precision

Contact Info

Product

Resources

About