The massive size of modern neural networks has motivated substantial recent interest in network quantization, especially low-bit quantization. We introduce Stochastic Markov Gradient Descent (SMGD), a discrete optimization method applicable to training quantized networks. SMGD algorithm is designed for settings where memory highly constrained during training. provide theoretical guarantees perf...

The main task of drilling the formation by the bit is done via scraping, crushing and grinding. Discharging the fluid from the bit nozzles is done with high pressure which assists to break the rocks. Different parameters affect the bit selection and design for each drilling formation, but the most important one is drillability that depends on hardness of the formation. In this paper, the design...