A new approach of a memory efficient huffman tree representation technique

Sultana, Zinnia; Akter, Sultana

doi:10.1109/iciev.2012.6317482

Cited by 3 publications

(3 citation statements)

References 10 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For the sake of completeness, it is worth noting that, when needed, the overhead can be still reduced at the expense of decoding speed, as there are methods storing a Huffman code for n symbols using at most 10.75n − 3 bits [47].…”

Section: A Huffman Address Map Compressionmentioning

confidence: 99%

Efficient and Compact Representations of Deep Neural Networks via Entropy Coding

Marinò,

Furia,

Malchiodi

et al. 2023

IEEE Access

View full text Add to dashboard Cite

Matrix operations are nowadays central in many Machine Learning techniques, including in particular Deep Neural Networks (DNNs), whose core of any inference is represented by a sequence of dot product operations. An increasingly emerging problem is how to efficiently engineer their storage and operations. In this article we propose two new lossless compression schemes for real-valued matrices, supporting efficient vector-matrix multiplications in the compressed format, and specifically suitable for DNNs compression. Exploiting several recent studies that use weight pruning and quantization techniques to reduce the complexity of DNN inference, our schemes are expressly designed to benefit from both, that is from input matrices characterized by low entropy. In particular, our solutions are able to take advantage from the depth of the model, and the deeper the model, the higher the efficiency. Moreover, we derived space upper bounds for both variants in terms of the source entropy. Experiments show that our tools favourably compare in terms of energy and space efficiency against state-of-the-art matrix compression approaches, including Compressed Linear Algebra (CLA) and Compressed Shared Elements Row (CSER), the latter explicitly proposed in the context of DNN compression.

show abstract

Section: A Huffman Address Map Compressionmentioning

confidence: 99%

Efficient and Compact Representations of Deep Neural Networks via Entropy Coding

Marinò,

Furia,

Malchiodi

et al. 2023

IEEE Access

View full text Add to dashboard Cite

show abstract

“…Butinreference [1], the author zinnia et al proposed the bit representation of Huffman tree, where the authors use some terminology namely the conception of "circular leaf node", "node with 2 internal son nodes" and give new concepts of "upper leaf node", "external limb", "internal limb", "antenna limb" [1] to solve critical limitations of existing works and they were promising to solve some existing problems.…”

Section: Literature Reviewmentioning

confidence: 99%

“…Then a massive change in algorithm would be needed in that case. **In reference [1], the best case will be occurred when only the right most branch of the tree is expanded which is not a very frequent occurrence. In other cases redundant bits for saving the external limb and each upper leaf node will be required.…”

Section: Figure 3: Line Diagram Showing Required Number Of Bits For Dmentioning

confidence: 99%

An Improved Decoding Technique for Efficient Huffman Coding

Mustafa¹

2017

JCSAIT

View full text Add to dashboard Cite

In today's world storing numerous data or information efficiently is a significant issue because of limited storage. Every moment we need to transfer large volume of information soundly and correctly. A storage space/memory storage required to store this information is. The way of communication which is also constricted with confined communication lines. These incompetence drives to need of data compression. Compression makes a reduction in memory storage, data transmission time, communication bandwidth and the ultimate cost savings also. To compress data efficiently and effectively one of the most popular and widely used techniques is Huffman compression. It is a lossless compression technique that enables the restoration of a file to its authentic/key state, having not to loss of a single bit of data when the file is uncompressed. It will be more efficient by reducing the memory requirements for Huffman tree. This article aimed at reducing the tree size of Huffman coding and also explored a newly memory efficient technique to store Huffman tree. As a consequence we also designed an encoding and decoding algorithm. Our proposed technique is much more efficient than most of the/all other existing techniques where to represent the Huffman tree structure total memory requirements are 9n-2 bits for the worst case, average and also for the best case. The results obtained significant improvements over previous/existing work. 2

show abstract