字幕列表 影片播放 列印所有字幕 列印翻譯字幕 列印英文字幕 The main video is talking about a genetic breeding model 主影片是在探討遺傳育種模型 of how to make machines learn. 如何使機器學習 This method is simpler to explain or just show. 這種演算法比較容易解釋與呈現 Here is a machine learning to walk or play Mario or jump really high. 這裡有小機器人學著走路 或學玩瑪莉歐 或學跳高高 A genetic code is an older code but it still checks out 遺傳算法已經過時 但仍有用 and I personally suspect in the future genetic models will have a resurgence as 而且我個人認為將來遺傳算法將會崛起 compute power approaches crazy pants. 一旦電腦計算能力接近嚇死人的地步 However, the current hotness is deep learning and recursive neural networks and 然而 現今最熱門的技術是深度學習和遞歸神經網絡 that is where the linear algebra really increases 這也是為什麼線性代數的重要性不斷提升 and explainability in a brief video really decreases. 而用簡短的影片來解釋的可行性愈來愈低 But if I had to kind of explain how they work 但要我用小影片來稍微解釋它們如何運作的 in a footnote, just for the record, it's like this: 只是為了點明 那會像這樣: No infinite warehouse. Just one student. 沒有無限大的教室 只有一個學生 Teacher Bot has the same test, but this time Builder Bot is 'Dial Adjustment Bot' 教師演算法的考卷不變 但現在「打造者演算法」是「調節員演算法」 where each dial is how sensitive one connection in the student bot's head is. 每一個旋鈕決定學生演算法腦中 每個神經連結的敏感度 There's a lot of connections in its head so a lot of dials. 在它腦中有非常多神經連結 所以有非常多旋鈕 A LOT, a lot. 超級 超級多 Teacher Bot shows Student Bot a photo and Dial Adjustment Bot adjusts that dial 教師給學生一張圖片 同時調節員調整旋鈕 stronger or weaker to get Student Bot closer to the answer. 使之更強或更弱 讓學生更接近正確答案 It's a bit like adjusting the dial on a radio. Is that still a thing? Do cars have radios still? 很像是在調整收音機的頻率 I don't know, anyway. You might not know the exact frequency of the station 現在還有收音機嗎? 汽車還有這東西嗎? but you can tell if your getting closer or further away. 我不知道 算了 It's like that but with a hundred thousand dials and a lot of math, 你應該不知道你要聽的電台的準確頻率 and that's just for one test question. 但你能夠分辨出是否接近要聽的電台 When Teacher Bot introduces the next photo, 這點和這種演算法很像 但有十萬以上的旋鈕 跟一大堆數學 Dial Adjustment Bot needs to adjust all the dials so that Student Bot can answer both questions. 再者 這只是針對一個問題而已 As the test gets longer, 當教師拿出下一張圖片 this becomes an insane amount of math and fine tuning for Dial Adjustment Bot. 調節員必須去調整所有的旋鈕 使學生能給出兩個問題的答案 But when it's done there's a student bot who 隨著測試愈來愈長 can do a pretty good job at recognizing new photos, 數學運算量和微調次數愈來愈龐大 though still suffers from some of the problems mentioned in the main video. 但當完成時 學生演算法 Anyway, that the most babies' first introduction to neural networks you will ever hear. 就能將未見過的圖片辨識得非常好 If it sounds interesting to you and you like math and code, 雖然仍會遭遇一些問題 在主影片有提到 go dig into the details; 總之 這是你能聽到最簡單的神經網絡之簡介 machines that learn are the future of everything. 如果你很感興趣 而且你喜歡數學跟寫程式 Maybe, quite literally, the future of everything, 去挖掘更多知識吧 and given what we've put them through, may the bots have mercy on us all. 可以學習的機器是未來的一切
B1 中級 中文 演算法 學生 神經 教師 數學 電台 機器如何*真正*學習。 [腳註] (How Machines *Really* Learn. [Footnote]) 16 0 林宜悉 發佈於 2021 年 01 月 14 日 更多分享 分享 收藏 回報 影片單字