作者查詢 / goldflower

總覽項目：發文 | 留言 | 暱稱

作者 goldflower 在 PTT [ DataScience ] 看板的留言(推文), 共366則

限定看板：DataScience

看板排序：

全部Gossiping22625DigiCurrency5639RockMetal691Grad-ProbAsk642graduate610PC_Shopping453Soft_Job415Option379DataScience366HardwareSale340NCTU_TALK276Python273Tennis254LoL181C_Chat180Physics146Federer114HatePolitics112Headphone112NCKU_PHY_T-T105joke96Tech_Job96CMWang91PublicIssue79GraduateCram70Mix_Match68guitar59Stock52Salary49WomenTalk47MobileComm42WorldCup42car35YOLO35HSNU_112434NBA32Pedal28SkiSnowboard28NTU27PushDoll27Vocal27sex26HotBloodYuan23NHSH13th30523tennis_life23MenTalk19NCTU-Teacher19SMSlife19TigerBlue19Tobacco19HwangYih18Viator96Chia18DC17Kaohsiung17StupidClown17Baseball16GIEE_BASKET16DreamTheater15fashion14ForeignEX14IA14MP3-player14specialman14Bitcoiners13BuyTogether13FJU_Psy9813PublicServan13Trading13Football11HsinPu11KOTDFansClub11L_TalkandCha11Oversea_Job11AfterPhD10GO10Math10Beauty9CrossStrait9DummyHistory9home-sale9Olympics_ISG9Tainan9Teacher9WOW9bag8MuscleBeach8Statistics8TKU_EW94B8historia7MacShop7media-chaos7NKFUST-CCE907TaichungBun7Golden-Award6NTUT_EE491A6studyteacher6Suckcomic6TTU-Transfer6Badminton5biker5Boy-Girl5Brand5Broker5C_and_CPP5CSMU-HSA965CVS5Employee5FHSH-89-3165GossipPicket5HSNU_10375HSNU_10955IntlShopping5KS96-3115KS97-3055LTK5mobilesales5NCHU-FS995nCoV20195NCYU_DVM_985NTUST-ECE5NUU_ER5PttHistory5SSSH-16th-Fk5SuperStarAve5TKU_EE_92C5X-Japan5ASHS-95RN4ChangHua4consumer4FCU_EE97A4Flute4HCU4HSNU_10984MAC4MdnCNhistory4Mongolia4NCYU_Fst_984NTHU-EE-CAPT4NTHU_STAT944NTPU-STAT964NTUST_Talk4NUU_CSIE4optical4PingTung4regimen4SCU_Talk4Sucknovels4SweetyAnts4TheWall4TOEIC4TTU-EE994TW-history4Warfare4balaz3CFantasy3cksh79th013CSMU-ROCK3DirectSales3DistantLove3dlsh-7th-3033Doraemon3e-seller3e-shopping3Hate3HCHS603093HK-movie3HOT_Game3HSNU_8203Instant_Mess3J-PopStation3Japan_Travel3Militarylife3MUSTMIS3nb-shopping3NFU3ONE_PIECE3paranormal3piano3PokeMon3SouthPark3Test3Wanhua3Windmill3About_Life2Ame_Casual2Anti-Fake2Anti-ramp2AntiVirus2Aves2Cancer2ck50th3052ck57th3142ckmc2CSMUdancepub2CT24th3372Ecophilia2ESP2Espannol2FCU_EE00B2Finance2FLAT_CLUB2HSNU_10082Instant_Food2IronMaiden2Jacky_Woo2KTV2Little-Games2Management2MartialArts2medstudent2MGL-history2Miaoli2movie2Mudran2Nantou2NCCU05_GIDS2NCKU_EARTH982NCTU_HSS982NCTU_ME_PES2NTNU-SLM952NTUCivilism2NTUT_ME495A2PokemonGO2pts2R_Language2ScienceNote2TAKESHI2TAXI2TFSHS66th3072TKU_IHM2toberich2transgender2USC2WoodworkDIY2WorkinChina2YiGo3112YuanLi3042Aboriginal1Actuary1allergy1AndyLau1Anti-Cancer1ASIA-uni1Audiophile1Azumanga1BabyMother1bass1Bioindustry1bitter1Board-Police1Brit-pop1Buddhism1CareerPlan1CATCH1Chat82gether1Christianity1ChungLi1ck50th3231ck_17_3011CKCB1cksh83rd3141cksh83rd3171cksh84th3101cksh84th3121cksh85th3011CMS_97_S3F1CMU_chorus1CodeJob1CompBook1CPU_FC7611creditcard1CSMU-D881CSMU-D971CSMU-MT991CSU1CYSH97Y3181DaBones1DCSH_7th_3081Deutsch1DivingSport1documentary1dog1DPP1Drink1Drum1DYU1Emergency1Eng-Class1eSports1Examination1facelift1fastfood1FCU-INS93B1Feminism1FITNESS1FJU-CH-2007B1FJU-Family1FJU-Stat95A1FJU_Psy951France1FSHS-95-3081Fund1gallantry1GameDesign1Gintama1haiku1hardware1HatandCap1HCHS593051HCHS593131HCHS603121HCSH_10th3121Health1Hip-Hop1HLHS5thU1Hsinchu1HSNU_10271HSNU_10821HSNU_11061HSNU_11091HSNU_11151HSNU_11491HSNU_9481Hualien1Hunter1I-Lan1ID_Multi1ILSH-973091Ingress1JinYong1JOKER1JP_Custom1JP_Entertain1KingofPop1Kings1KMT1kochikame1KS90-3031KS95-3101KS96-3051KS96-3181KS97-3101KS97-3141KUAS1Lab7201LAND_BASKET1Laser_eye1Lawyer1LD_IM93-11LifeSci_961LightNovel1LinKou1LittleNew1LTU1Malaysia1Marginalman1marvel1MayDay1Media-work1MH1Miracle1ModernRock1Nadal1NARUTO1Navy1NCCU06_BA1NCCU07_LAND1NCCU07_SW1NCCU08_Ethno1NCCU08_LAND1NCCU_RNR1NCCU_SEED1NCHU-FS981NCHU-MEbaske1NCHU-Wind1NCKU-physiol1NCU97ME-B1NCUFingrad031NCYU_AFL_961NDHU-His1021NDHU-His961NDHU-LS991NDHU-phy951NDHU_FIN981NDHU_MSE931NDMC-M1061NEHS19th41NetSecurity1NHK1NHLUE-ILT1Nirvana1NIU-ECE94b1NPTU_CaC1NSYSU1NTHU_CKCM1NTPU-CSIE931NTUBSE-B-971NTUE-DC991NTUE-TGD971NTUE_NSE1001NTUE_Nse1011NTUE_Nse1021NTUEE_BL5271NTUEE_Lab2071NTUEE_LAB5061NTUHistory021NTUmed911NTUST-TX-B921NTUST-TX-B951NUK_EE100A1NUTN_MS991Orzhong97cl1PCSH92-3031PCSH97_3131PeopleSeries1pharmacist1PlayStation1PostRock1Printer3D1Programming1PTHS95-3101Ptt-Press1PttCard1PttLifeLaw1R963300XX1rent-exp1RSSH90_3021SCUG1SENIORHIGH1SFHS1shoes1SlamDunk1soul1SSSH-09th1141Steam1Storage_Zone1studyabroad1StudyGroup1Sub_DigiLife1SW_Job1TaitungH92-11TallClub1TCFSHvolley1TFSHS68th3211TFSHS69th3231THU-CHE931THU-CHE961THU-P-Softbo1THUAS_911THUIM-6th1THULAW1TKU_Talk1Transformers1trumpet1TuTsau1TW-language1TypeMoon1TYSH49-3021UTAH-JAZZ1VET_961VictoryYouth1VideoCard1WuLing50-3171YMU_BME951YoungDotx31YP85-3121Zastrology1ZLSH1<< 收起看板(468)

首頁

尾頁

[問題] 請問關於multi label 的paper關鍵字

[ DataScience ]4 留言, 推噓總分: +2

作者: filialpiety - 發表於 2022/05/24 19:40(3年前)

2^F推goldflower: hierarchical multi-label classification05/25 01:59

3^F推goldflower: hmm這我不確定你可以把這些關鍵字後面加上github05/26 12:51

4^F→goldflower: 或paper with code看看有實際應用在什麼問題上05/26 12:52

[問題] 標記的圖片數據集,樣本個體之間差異性太小

[ DataScience ]41 留言, 推噓總分: +6

作者: ruthertw - 發表於 2022/05/23 00:21(3年前)

21^F推goldflower: 但你拿來judge的若都是那個test set你搞這堆有意義嗎05/25 02:02

22^F→goldflower: 你至少要拿更符合母體的test set才能去評斷你做的事情05/25 02:03

23^F→goldflower: 有沒有用05/25 02:03

Re: [問題] 27歲轉職資料科學相關工作

[ DataScience ]20 留言, 推噓總分: +6

作者: kimoji - 發表於 2022/03/19 12:50(3年前)

20^F推goldflower: 去加拿大混個短期學歷直接找當地工作03/20 16:36

[問題] 模型及工具挑選

[ DataScience ]7 留言, 推噓總分: +4

作者: ccbruce - 發表於 2022/03/15 20:27(3年前)

1^F→goldflower: 相同的東西隨便弄個feature matching就好了吧03/15 23:38

[問題] 不同硬體下訓練，驗證組表現出現差異

[ DataScience ]4 留言, 推噓總分: +1

作者: subgn - 發表於 2022/03/13 19:04(3年前)

3^F推goldflower: 出包的可能環節很多啊 cuda/tf/onnx/m1 blabla03/15 23:37

[問題] 25歲轉職資料科學

[ DataScience ]39 留言, 推噓總分: +12

作者: Trumpet01017 - 發表於 2021/12/23 15:57(4年前)

39^F推goldflower: 出國嚕可以修資工學分的學歷02/21 23:37

[問題] Keras版本不同對訓練結果的影響

[ DataScience ]9 留言, 推噓總分: +1

作者: ArthurJack - 發表於 2020/12/16 14:42(5年前)

5^F推goldflower: 建議是要用keras請通通改tf.keras12/17 12:50

6^F→goldflower: keras已經不會獨立維護了12/17 12:51

7^F→goldflower: 你先用tf.keras不行再說12/17 12:51

8^F→goldflower: 另外據我經驗tf1請用1.14 不然就改用tf2了 1.15蠻雷12/18 18:12

[討論] Dense layer 的特性

[ DataScience ]13 留言, 推噓總分: +5

作者: acctouhou - 發表於 2020/12/13 21:43(5年前)

2^F推goldflower: 應該沒啥假設吧大多用orthogonal initialization加12/14 17:41

3^F→goldflower: 上loss constraint去達成你想要的正交效果但我有用12/14 17:41

4^F→goldflower: 的都是去要求layer內的weight正交 output間正交或許12/14 17:41

5^F→goldflower: 可以參考看看orthogonal low rank embedding這篇？12/14 17:41

9^F推goldflower: 跟輸入正交沒啥關係12/15 09:52

[問題] Edge AI 落地實作班

[ DataScience ]5 留言, 推噓總分: +1

作者: jienfong - 發表於 2020/11/25 21:59(5年前)

1^F推goldflower: edge的硬體很多種啊最多資源就手機了 tflite現在比11/27 18:14

2^F→goldflower: 以前輕鬆不少若要training目前只能caffe或自己刻手11/27 18:14

3^F→goldflower: 機以外應該更多問題我覺得沒什麼人會教欸11/27 18:14

[問題] 需要每一輪(epoch)都餵入整個資料集嗎？

[ DataScience ]10 留言, 推噓總分: +4

作者: art1 - 發表於 2020/11/10 18:33(5年前)

8^F推goldflower: 當然沒這規定時間不夠就時間不夠的訓練法囉不然你11/13 15:15

9^F→goldflower: 要做decay之類的都不能做或者可以先整理一次資料把11/13 15:15

10^F→goldflower: 相似的排到下個epoch再訓練就好11/13 15:15

首頁

尾頁