دانلود کتاب الگوریتم های یادگیری ماشین تطبیقی ​​با پایتون

This book presents several categories of streaming data problems that have significant value in machine learning, data visualization, and data analytics. It offers many adaptive algorithms for solving these problems on streaming data vectors or matrices. Complex neural network-based applications are commonplace and computing power is growing exponentially, so why do we need adaptive computation? Adaptive algorithms are critical in environments where the data volume is large, data has high dimensions, data is time-varying and has changing underlying statistics, and we do not have sufficient storage, computing power, and bandwidth to process the data with low latency. One such environment is computation on edge devices. Due to the rapid proliferation of billions of devices at the cellular edge and the exponential growth of machine learning and data analytics applications on these devices, there is an urgent need to manage the following on these devices: • Power usage for computation at scale • Non-stationarity of inputs and drift of the incoming data • Latency of computation on devices • Memories and bandwidth of devices The 2021 Gartner report on Edge computation [Stratus Technologies, 2021] suggests that device-based computation growth propelled by the adoption of cloud and 5G will require us to prioritize and build a framework for edge computation.