程序代写代做代考 deep learning 題目:Approximate Computing for Deep Learning in TensorFlow

題目:Approximate Computing for Deep Learning in TensorFlow

內容:MobileNet是個新提出的基於approximate computing的deep learning framework, 我們想測試比起老牌的deep learning演算法 (如vggnet), 是否達到了降低少許accuracy, 同時卻能達到大量提升速度的效果; 此外, MobileNet還有兩個參數 – width multiplier和resolution multiplier可更近一步地實現approximate computing的概念, 我們將探索這兩個參數導致的accuracy與computation time的trade-off關係

預期結論:MobileNet確實能在損失少許accuracy的情形下, 大幅提升速度, 故其非常適合作為手機app可連接的API, 因為他速度快, 能耗低

Posted in Uncategorized

Leave a Reply

Your email address will not be published. Required fields are marked *