本文旨在探討 機器學習分類方法 ”支持向量機 ”及”Softmax ”在圖像辨識 上的訓練結果比較，也研讀當今常用優化算法熟知損失函數對參更新影響關係，兩個模型訓練將以蝴蝶圖片進行實作。 經由線上開放的圖片庫網站，取得 8214 張共五類蝴蝶，並自製成數據樣 本集，分別帶入兩個 訓練 模型，觀察兩者訓練時間及準確率在迭代結果 上分析 比較 。而後再進一步 探討影響訓練結果的原因 ，在數據預處理 上，看 不同的數據庫量是否影響 訓練或驗證準確度。最後將模型結果對測試集 進行預測，觀察準確率分析探討影響結果的因素。;The purpose of this thesis is to explore the training resul ts of two deep learning models ：(1) Support Vector Machine ；(2) Softmax Classifier in image recognition, and study the influence of loss functions on the iterative parameters . We demonstrate the results of these two models by use of the image s of butterflies. There are five types of butterflies with 8214 pictures obtained through the onlin e database website. We use these pictures for the files of data samples to two deep learning models, and observe the training time and accuracy. Next, we analyze the fitting situation to the itera tive results. Finally, we give the reason s why the performan ce of these two models are not ideal. Therefore, we are able to improve the performance by fixing the datasets .
