Mutazhakmi Direct

[ \Delta! \textAcc(\mathbfC) \approx \sum_i=1^N \left( \lambda^(b) i \cdot (b \textref-b_i) + \lambda^(p)_i \cdot p_i \right) ]

The MH framework builds on these ideas but between model scaling, hardware mapping, and energy budgeting, enabling automatic generation of Pareto‑optimal solutions. mutazhakmi

Hakmi has served as a mentor for events like Startup Weekend Homs , helping the next generation of Syrian entrepreneurs and developers bring their ideas to life. Impact on the Tech Ecosystem [ \Delta

| Task | Dataset | Baseline Model (FP32) | |------|---------|-----------------------| | Image Classification | CIFAR‑10 | ResNet‑20 | | Image Classification | ImageNet‑Mini (100 classes) | MobileNet‑V2 | | Speech Command Recognition | SpeechCommands‑V2 | 1‑D ConvNet | | Human Activity Recognition | UCI HAR | LSTM‑2×64 | Impact on the Tech Ecosystem | Task |

When (f_i = 1), the scheduler merges the computation graph of (L_i) and (L_i+1) into a single kernel, reducing memory traffic. The feasibility check uses a derived from the TVM tensor‑intrinsic capabilities of the target device.

M. A. Mutaz, L. Hakmi, S. R. Bennett, J. K. Liu Department of Computer Science, Global Institute of Technology, Singapore Correspondence: m.mutaz@git.edu