• Poonam Domkundwar
• Sugandha Nandedkar

Federated learning licenses various social affairs to together train a profound learning model on their joined data, with no of the individuals revealing their local data to an incorporated worker accordingly supporting privacy conservation. This kind of security sparing aggregate learning, regardless, comes to the detriment of an important correspondence overhead during getting ready. To address this issue, a couple of weight methods have been proposed in the scattered getting ready composing that can lessen the proportion of required correspondence by up to three huge degrees. These current procedures, regardless, are simply of obliged utility in the federated learning setting, as they either simply pack the upstream correspondence from the clients to the worker (leaving the downstream correspondence uncompressed) or simply perform well under meager conditions, for instance, i.i.d. movement of the client data, which customarily can't be found in federated learning. In this article, we propose Vector Compression Technique (VCT), another pressure structure that is expressly planned to meet the necessities of the federated learning condition. VCT expands the current weight system of top-k tendency sparsification with a novel instrument to engage downstream weight and ideal encoding of the weight refreshes.

"Federated learning using Vector Compression Techniques", International Journal of Emerging Technologies and Innovative Research (www.jetir.org), ISSN:2349-5162, Vol.7, Issue 11, page no.136-140, November-2020, Available :http://www.jetir.org/papers/JETIR2011317.pdf

"Federated learning using Vector Compression Techniques", International Journal of Emerging Technologies and Innovative Research (www.jetir.org | UGC and issn Approved), ISSN:2349-5162, Vol.7, Issue 11, page no. pp136-140, November-2020, Available at : http://www.jetir.org/papers/JETIR2011317.pdf