Paper review on predicting the deep network generalization gap from margin-distribution statistics across hidden layers, fed into a simple linear estimator.
A practical introduction to mixed-integer programming for ML engineers, covering branch-and-bound, presolve and cutting planes with hands-on examples in CVXPY.
Our paper on scoring neuron importance with mixed-integer programming to prune neural networks without fine-tuning, using ReLU bounds and a sparsity-driven objective.
Walkthrough of PackNet, a parameter-isolation method that fits multiple continual learning tasks into one network through iterative magnitude pruning and retraining.
Paper review of stabilizing GAN training with random projections, using an ensemble of discriminators on fixed low-dimensional views to fight diminishing gradients.
My time at Raisa Egypt forecasting oil and gas production with machine learning, modernizing legacy data collection tooling and running a company-wide hackathon.
Engineering OpenCV image features such as simplicity, sharpness, contrast and face counts to boost a LightGBM ad-demand model for the Avito Kaggle competition.
My journey at Microsoft Research Lab Cairo working on Skype Translator: dialectal Arabic NMT, gender-aware translation, speech recognition and multilingual chatbots.
