Adaptive Exponential Weight Discretization for Compact and energy-efficient CNN Inference
Bibliographic record
Abstract
Convolutional Neural Networks (CNN) compression and optimization techniques for the inference process attract attention due to CNN’s broad range of applications in many fields. Implementing CNNs for embedded deceives, however, faces great challenges due to the devices’ limited hardware resources. Recently, real-time pplications based on complex CNNs require optimization and compression techniques to make CNNs run faster in compact and energyefficient embedded devices. CNN odel optimization and compression include pruning and weight quantization. Conventional pruning techniques often suffer from poor compression ratio, so they still have a large number of float point weights leading to embedded devices with excessive hardware cost or slow inference speed. Conventional quantization techniques often result in an unacceptable loss in inference accuracy. To overcome the above problems, the proposed method introduces an adaptive exponential weight discretization process, which selects discretization parameters based on the sensitivity and the number of the weights in each layer. Unlike many previous compression methods, the proposed method does not require any retraining process nor any fine-tuning methods. The proposed weight discretization has been evaluated using the VGG16 CNN model with the ImageNet dataset for classification. The evaluation demonstrated that it saves 48 % of computation for fully connected layers and it reduces the overall weight memory size from 528 MB to 44.24 MB – a reduction of 11.9 times. We also demonstrate that the optimized CNN loses negligible loss – a loss of only .43% in the top-5 accuracy and a loss of only 1.44% in the top-1 accuracy.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".