https://chih-sheng-huang821.medium.com/稍微削減decision-tree-和-stump做base-model的boosting-model-沒有什麼優勢不優勢-以我的觀點來看-stump做base-model的boosting-a926bf02af7b2020-04-24monthly0.5 https://chih-sheng-huang821.medium.com/你好-boosting利用ensemble的方式可以同時減少bias和variance-利用多個weak-classifier達到stronger-15be71070b632020-04-23monthly0.5 https://chih-sheng-huang821.medium.com/因為範例我是隨便給予一個值-不是真的跑模型得到的結果-b3c0a31ab3132020-04-23monthly0.5 https://chih-sheng-huang821.medium.com/precision-所有被檢測為目標-但-正確分類為目標-的比例-4b6b8251e8382020-04-23monthly0.5 https://chih-sheng-huang821.medium.com/其實tree本身就是一種bagging-e91a4652cdbf2020-04-23monthly0.5 https://chih-sheng-huang821.medium.com/感謝你發問-我用的是-anchor-5-c-number-of-283ef64c493a2020-04-19monthly0.5 https://chih-sheng-huang821.medium.com/一樣用下面表格的方式算-只是資料從五筆改成改成三筆-但計算三筆資料的相關係數應該沒什麼意義-ef1eece1c2ab2020-04-14monthly0.5 https://chih-sheng-huang821.medium.com/深度學習鮮少在用cv做參數搜尋-因為上一篇回答已經達過時間會很長不太可能把幾百萬組組合都跑過還找參數-就算是nas系列搜尋模型已經夠可怕-也不可能所有狀況都跑過-d1c32a6ac29e2020-04-14monthly0.5 https://chih-sheng-huang821.medium.com/感謝你的留言-留言寫-然而你文章中寫的是先對每個類別中的不同樣本的概率分佈進行log求和-然後再將每個類別的ce求和得到最終ce-loss-這樣是不對的-應該是指下面這段吧-c1bd9ccf9f9b2020-03-27monthly0.7 https://chih-sheng-huang821.medium.com/tf的文件數是用2-spam-and-genuine-對-ec924d8c4f1c2020-06-02monthly0.5 https://chih-sheng-huang821.medium.com/謝謝拉-當初我寫這篇我自己都看到眼花-add23202a9f52020-05-28monthly0.5 https://chih-sheng-huang821.medium.com/是-林進燈老師沒錯-有想要找我討論的-你可以丟我linkedin站內信喔-f628e05943152020-05-28monthly0.7 https://chih-sheng-huang821.medium.com/我沒研究過kmo做什麼事情-但九個變量-到四個累積解釋變異才71-背後的意義代表資料變數間的變異量基本上都很高且共變異量可能不高-變數間可能快獨立了-所以有可能很難找到資料背後的因子-不合適做因素分-2e135d33da402020-05-18monthly0.5 https://chih-sheng-huang821.medium.com/route-894d7d0398472020-05-18monthly0.5 https://chih-sheng-huang821.medium.com/那換個角度思考-你輸入的特徵數量是否因為dummy-variable導致特徵數太多-如果太多可以建議你類別資料先用1-2-3-4-的方式取代dummy-variable的one-hot-965b8ddd2c812020-05-16monthly0.5 https://chih-sheng-huang821.medium.com/我沒有用spss-但對數概似值應該是指上式-所以值等於0表示p-1-p-1或beta-xi-0-所以p-0-5-也就是成功的機率跟失敗的機率都是0-5-所以等於模型完全無鑒別度-所以有可能beta-43710f7bd0f32020-05-15monthly0.5 https://chih-sheng-huang821.medium.com/文中講的ml-dl之差異是否只是監督與非監督的差異呢-b0d40e29af682020-07-28monthly0.7 https://chih-sheng-huang821.medium.com/k-k-nh-nk-is-for-the-standard-conv-a80e5ff1cfd82020-07-06monthly0.5 https://chih-sheng-huang821.medium.com/in-my-opinion-it-seems-like-2ed914bd569d2020-08-31monthly0.5 https://chih-sheng-huang821.medium.com/我這邊要提的是說數據量的不平衡不一定會對map判斷造成影響-主要影響是來自資料的代表性-也就是學習時候的資料是否足以代表整個母體-16f5cc216d162020-08-20monthly0.5 https://chih-sheng-huang821.medium.com/我看email你應該是說這段-ans-機率越隨機-可能一下成績高一下成績低-的情況-訊息量比較大-應該是越小吧-3e6f0e64df702020-08-20monthly0.7 https://chih-sheng-huang821.medium.com/這邊需要強調一點-數據量的差異不一定對決策有影響-可以查一下bias-variance-c1486e84fe642020-08-20monthly0.5 https://chih-sheng-huang821.medium.com/我沒有看過一般教科書怎麼介紹這部分-cad3543c3a422020-10-14monthly0.5 https://chih-sheng-huang821.medium.com/你可以貼你的confusion-matrix和roc給我看一下嗎-e3fe056dd1742020-10-14monthly0.5 https://chih-sheng-huang821.medium.com/你試著畫roc看看-905c99d1ab242020-10-13monthly0.5 https://chih-sheng-huang821.medium.com/笨蛋-問題在data-abc60479a0532020-10-13monthly1.0 https://chih-sheng-huang821.medium.com/文章是-機器學習-統計方法-模型評估-驗證指標-validation-index-af119b77a9e32020-10-11monthly0.5 https://chih-sheng-huang821.medium.com/對-打錯了-感謝你-704e30ba3ca92020-09-15monthly0.5 https://chih-sheng-huang821.medium.com/文章內的-一開始有說yolo最後的tensor為s-s-b-5-c-以voc的例子來說輸出為7-7-2-5-20-7-7-30-a9e62b947bd72020-09-13monthly0.7 https://chih-sheng-huang821.medium.com/應該大陸也很多類似的文章你可以找看看參考-我是用我自己的理解來寫這些文章-公式推導是我自己從頭推一遍然後才寫的-所以有些部分會有打錯-靠大家細心閱讀幫我找出一下小bug-4fb0f857f4da2020-10-16monthly0.7 https://chih-sheng-huang821.medium.com/文章有寫-後面的圖例-事前機率我先設定每一類都是一樣的-感謝你-627b372a4aa12020-12-14monthly0.5 https://chih-sheng-huang821.medium.com/這邊有點tricky-因為只是套用這樣的手法來推論我們要的結果-78adc5a0a6ac2020-12-01monthly0.5 https://chih-sheng-huang821.medium.com/是我沒有寫清楚-如果a是參數-對a做偏微分-然後讓設定偏微分後的函數為0去求解a的參數解-這樣得到的a可能為極值-也就是求出來的a-可以讓var-ax-y-有極大或是極小值-40c10c7266ca2020-11-30monthly0.5 https://chih-sheng-huang821.medium.com/是-已修正-感謝-2166fa818a072020-11-30monthly0.5 https://chih-sheng-huang821.medium.com/about2026-03-06weekly1.0 https://chih-sheng-huang821.medium.com2026-03-06weekly1.0 https://chih-sheng-huang821.medium.com/yes-f7b9bf96d3962020-11-19monthly0.5 https://chih-sheng-huang821.medium.com/y-bar-跟-x-bar是資料的平均數-a7370e14b5652020-11-19monthly0.5 https://chih-sheng-huang821.medium.com/對阿-likelihood-function不是機率密度函數-嚴格一點定義-likelihood-function為在得知觀察值-observed-data-4f4c870f386e2020-12-25monthly0.7 https://chih-sheng-huang821.medium.com/深度學習warm-up策略在幹什麼-95d2b56a557f2020-12-23monthly1.0 https://chih-sheng-huang821.medium.com/d是-所有文件的總數-但也可能為-所有詞的總數-以範例來說-假設文件有100個-那d就是100-cf7f46e737892021-02-09monthly0.5 https://chih-sheng-huang821.medium.com/的確在deep-learning相關沒有用kernel-map這個用字-kernels在形容-in-channel-out-channel-k-k-我自己用kernel-8dbc4d9ce64c2021-02-03monthly0.5 https://chih-sheng-huang821.medium.com/嚴格來說kernel-map這個名詞出現在svm那種kernel-method才對-b6fb36c773fe2021-02-02monthly0.7 https://chih-sheng-huang821.medium.com/已修正-感謝-112bfb7e5bc82021-02-01monthly0.5 https://chih-sheng-huang821.medium.com/笨蛋-問題在data-2-從訊息理論來看資料的影響-9aa2e2b420c62023-10-27monthly1.0 https://chih-sheng-huang821.medium.com/是面積沒錯-下一段話開頭有寫面積-此範例函數我呈現的是連續函數-機率論的連續函數單點無機率-因此我才會寫機率密度函數值-但有個失誤-應該寫積分值或是機率值就好-感謝你-8e199d8acda32021-01-20monthly0.7 https://chih-sheng-huang821.medium.com/好的-38774e889b3e2021-02-27monthly0.5 https://chih-sheng-huang821.medium.com/torchvision都只有圖的轉換-如果你是做物件偵測或是segmenation則需要將標註也一起改-直接用torchvision就不合適了-9f673f3ce4042021-02-24monthly0.5 https://chih-sheng-huang821.medium.com/03-pytorch-dataaug-a712a7a7f55e2021-02-24monthly1.0 https://chih-sheng-huang821.medium.com/你指的是三個時間點2個dependent-variables-總共只有6個數字-如果是這樣數據不太足夠-算correlation不太適合-應該說數據量過小然後要做統計是沒有意義的-5f81103ec8122021-03-25monthly0.7 https://chih-sheng-huang821.medium.com/感謝你發現這個大錯誤-已修改內容-ac9a92d52b422021-03-15monthly0.5 https://chih-sheng-huang821.medium.com/深度學習-物件偵測yolo-cfg檔解讀-三-2021年-2faa5c19fd362021-04-22monthly1.0 https://chih-sheng-huang821.medium.com/okay-2fc97c65e6d62021-06-03monthly0.5 https://chih-sheng-huang821.medium.com/修改好了-感謝-9717de6cb42021-08-15monthly0.2 https://chih-sheng-huang821.medium.com/pytorch手把手實作-autoencoder-f5a048fcab5b2024-12-26monthly1.0 https://chih-sheng-huang821.medium.com/pytorch手把手實作-generative-adversarial-network-8adae9a3092b2021-08-21monthly1.0 https://chih-sheng-huang821.medium.com/x1-x2-xn-iid-n-mu-sigma-所以表示每個變數服從平均數是mu變異數是sigma的常態分布-每個變數的期望值都為mu-因此e-x1-mu-c6599faea2e22021-10-10monthly0.7 https://chih-sheng-huang821.medium.com/機器學習的統計基礎-深度學習背後的核心技術-530caf5ea7952022-02-20monthly1.0 https://chih-sheng-huang821.medium.com/讀者留言回覆-38c9f34f7db32022-01-04monthly0.5 https://chih-sheng-huang821.medium.com/這邊提到的還是粗略的hard-target-也就是-你提到的左機率是one-hot-encod-但深度學習還有soft-target-並非只能用hard-8fb5a6451df2021-12-31monthly0.7 https://chih-sheng-huang821.medium.com/還看不懂wasserstein-distance嗎-看看這篇-b3c33d4b9422023-12-06monthly1.0 https://chih-sheng-huang821.medium.com/指數-exponential-級數和極限定義-97bfa94769e52022-01-13monthly1.0 https://chih-sheng-huang821.medium.com/影像切割任務常用的指標-iou和dice-coefficient-3fcc1a89cd1c2022-01-07monthly1.0 https://chih-sheng-huang821.medium.com/對寫錯了-要改成負號-d4adf46d55922022-02-07monthly0.5 https://chih-sheng-huang821.medium.com/調頻連續波雷達-frequency-modulated-continuous-wave-fmcw-radars-mimo-radar-6ed0cfe1b5972022-05-02monthly1.0 https://chih-sheng-huang821.medium.com/調頻連續波雷達-frequency-modulated-continuous-wave-fmcw-radars-測距-測速原理-3-3-5e19a7a9e12022-05-02monthly1.0 https://chih-sheng-huang821.medium.com/調頻連續波雷達-frequency-modulated-continuous-wave-fmcw-radars-測距-測速原理-1-2-cc6ac4b501b42022-09-07monthly1.0 https://chih-sheng-huang821.medium.com/調頻連續波雷達-frequency-modulated-continuous-wave-fmcw-radars-測距-測速原理-2-3-981e4706fc9d2022-05-02monthly1.0 https://chih-sheng-huang821.medium.com/謝謝你-我看懂你的意思-c342641e12232022-05-20monthly0.5 https://chih-sheng-huang821.medium.com/對-在focal-loss文章是寫alpha-t-但你實際上-α-balanced-ce-loss是定義alpha而已-focal-loss寫成alpha-t的原因也是是-方便記-而已-論文的3-1-7e68da21fedb2022-05-16monthly0.7 https://chih-sheng-huang821.medium.com/笨蛋-問題在data-3-probably-approximately-correct-pac-ec70e62c4bb92022-06-05monthly1.0 https://chih-sheng-huang821.medium.com/面對未知類別你說random-seed會造成巨大的差異-這個問題就是這筆資料-未曾-部分-出現在訓練資料內-e89b6ad6fcbe2022-06-21monthly0.7 https://chih-sheng-huang821.medium.com/如果你的訓練集是貓狗-然後要辨識出狐狸和熊-不可能-37a0d77dc97c2022-06-20monthly0.5 https://chih-sheng-huang821.medium.com/一般連續變數通常是就直接進演算法內-72847196db202022-07-02monthly0.7 https://chih-sheng-huang821.medium.com/科技財知道訪問-ai-工程師到底在幹麻-f99becbe858a2022-06-28monthly1.0 https://chih-sheng-huang821.medium.com/你好-a41bc1ac89aa2022-07-15monthly0.7 https://chih-sheng-huang821.medium.com/0-5-1-5-4-5-是輸出層到隱藏層-c35484960bcd2022-07-11monthly0.5 https://chih-sheng-huang821.medium.com/建議類別編碼的方式即可-因為除非你編碼方式不是用one-hot-encoding方式-如果是用例如-mean-encoding就建議做標準化-e0ad97f3cee72022-08-01monthly0.5 https://chih-sheng-huang821.medium.com/看你用什麼演算法的偵測head-6aeccddc64482022-08-01monthly0.5 https://chih-sheng-huang821.medium.com/我以前做的是multiple-classifier-system中的fusion算法-我看別人介紹也會稱multiple-classifier-system為ensemble-a4cd89533e522022-08-26monthly0.5 https://chih-sheng-huang821.medium.com/這句話我可能昏頭寫錯了-512e048d76d32022-09-07monthly0.7 https://chih-sheng-huang821.medium.com/ai模型壓縮技術-量化-quantization-9665051283652022-09-04monthly1.0 https://chih-sheng-huang821.medium.com/如果是最一般沒有變形的話-答案-是-8d0e31d21f4f2022-09-16monthly0.5 https://chih-sheng-huang821.medium.com/hi-因為我有寫一本書-68b7287fd7fc2022-11-30monthly0.5 https://chih-sheng-huang821.medium.com/可以喔-ec75b31875f72023-01-09monthly0.5 https://chih-sheng-huang821.medium.com/對耶-我寫錯了-b7e3bdb6dcdc2023-02-13monthly0.5 https://chih-sheng-huang821.medium.com/ml-通常適合處理數位的-結構化資料-類似excel那種表格資料-每一欄都是一個特徵值-例如身高體重這樣的數字-4d3cbbe2ac2f2023-03-10monthly0.7 https://chih-sheng-huang821.medium.com/hello-950d279cf3f62023-05-16monthly0.5 https://chih-sheng-huang821.medium.com/mastertalks課程推銷文-自己的課自己推銷-fc07d7d661fb2023-05-24monthly1.0 https://chih-sheng-huang821.medium.com/謝謝拉-不好意思這麼多天才看到訊息-非常感謝你找到錯誤-d0b637538c8f2023-08-03monthly0.5 https://chih-sheng-huang821.medium.com/我雖然寫期望值-但p-zk-1-xi-是任一實值可測函數-ln是凸函數也可以用jenson不等式-111589703c872023-08-21monthly0.5 https://chih-sheng-huang821.medium.com/公開演講-職涯講座-ai技能的培養提升-未來職涯競爭力-2023-10-14-d3e9380e58862023-09-18monthly1.0 https://chih-sheng-huang821.medium.com/我沒有用過耶-04f7f47429392023-10-19monthly0.5 https://chih-sheng-huang821.medium.com/對-是加號-謝錯了-7b13c8da245a2023-10-27monthly0.5 https://chih-sheng-huang821.medium.com/利用chatgpt和stable-diffusion進行ai故事創作-f0e834c012482023-11-20monthly1.0 https://chih-sheng-huang821.medium.com/簡易介紹卡爾曼濾波-kalman-filter-1b041e371fe62024-02-17monthly1.0 https://chih-sheng-huang821.medium.com/自駕車路徑規劃系列-3-笛卡爾座標系和frenet座標系相互轉換-bd0f1ac38ace2024-03-15monthly1.0 https://chih-sheng-huang821.medium.com/自駕車路徑規劃系列-1-lattice-planner之笛卡爾座標系和frenet座標系的關係-c14f03c24f2b2024-03-15monthly1.0 https://chih-sheng-huang821.medium.com/自駕車路徑規劃系列-2-frenet公式-8878b3bc71b92024-03-15monthly1.0 https://chih-sheng-huang821.medium.com/2024-05-22-mastertalks-線上課程平台線上直播職場分享-907c10425fe72024-05-16monthly1.0 https://chih-sheng-huang821.medium.com/1-是從斜率那邊推出來的-看斜率公式那段上面那張圖-0a63f4b55da52024-06-03monthly0.7 https://chih-sheng-huang821.medium.com/ppt投影片-不好意思沒注意到有流言-755d443a5ca52024-05-31monthly0.5 https://chih-sheng-huang821.medium.com/不好意思今天才看到訊息-8bca27913f862024-10-16monthly0.5 https://chih-sheng-huang821.medium.com/爬蟲擷取yt影片資訊和影片字幕-cc878b08a46a2025-01-09monthly1.0 https://chih-sheng-huang821.medium.com/利用openai-whisper來進行語音轉文字-51dbeca8fd602025-01-08monthly1.0 https://chih-sheng-huang821.medium.com/對-謝謝-我在改一下-ac258cf6c4762025-01-08monthly0.5 https://chih-sheng-huang821.medium.com/我知道最簡單的-單一層的perceptron的演算方式-本文圖1的w11-w12-w1d的這-一組-weights可想成是以一般統計學的迴歸係數估計方法所得出的-c57d13ee14a82025-01-06monthly0.5 https://chih-sheng-huang821.medium.com/你好-我也不是做雷達的-時間也有點久遠-所以回答可能有誤-你參考一下就好-8ea391839c4f2025-03-11monthly0.5 https://chih-sheng-huang821.medium.com/self-attention-自注意力機制-5737aa42072d2025-04-01monthly1.0 https://chih-sheng-huang821.medium.com/這個導數的確要以df-dx為主-d8c63a1c242d2025-03-27monthly0.5 https://chih-sheng-huang821.medium.com/基於numba的cuda-python介紹-ab44520c18362025-05-14monthly1.0 https://chih-sheng-huang821.medium.com/應該是程式碼的問題-我這邊請gpt幫忙驗證-他給的建議是下面的程式碼-e2104bb5fcba2025-08-06monthly0.5 https://chih-sheng-huang821.medium.com/維基寫的是標準rbf函數-所以會直接用φ-r-exp-ε-r-2-然後r-x-y-677efa43bbe22025-08-06monthly0.5 https://chih-sheng-huang821.medium.com/我現在才看到有問題-9116c431736a2025-08-20monthly0.5 https://chih-sheng-huang821.medium.com/驗證旋轉矩陣是正交矩陣-513b090ffb792025-12-04monthly0.7 https://chih-sheng-huang821.medium.com/你好-我其實不是做雷達的-我是做演算法的-1ebcb77f41ce2026-01-15monthly0.2 https://chih-sheng-huang821.medium.com/路徑規劃系列-b-spline-52b476adbabf2026-03-06daily0.5