投稿時間:2022-03-25 10:35:35 RSSフィード2022-03-25 10:00 分まとめ(38件)

カテゴリー等	サイト名等	記事タイトル・トレンドワード等	リンクURL	頻出ワード・要約等/検索ボリューム	登録日
TECH	Engadget Japanese	Google、Android版Spotifyでサードパーティのアプリ決済システムを試験へ	https://japanese.engadget.com/google-to-test-play-store-alternative-payment-system-with-spotify-000008781.html	spotify	2022-03-25 00:00:08
IT	ITmedia 総合記事一覧	[ITmedia ビジネスオンライン] ペット1頭当たりの1カ月の生活費用はいくら？　大型犬は“1万5000円未満”が最多	https://www.itmedia.co.jp/business/articles/2203/25/news076.html	itmedia	2022-03-25 09:50:00
IT	ITmedia 総合記事一覧	[ITmedia ビジネスオンライン] “コンビニの新商品”でチェックするモノ　「ドリンク」や「カップ麺」を抑えた1位は？	https://www.itmedia.co.jp/business/articles/2203/25/news075.html	itmedia	2022-03-25 09:35:00
IT	ITmedia 総合記事一覧	[ITmedia News] ウクライナは死亡したロシア兵の顔をAIで特定し、家族に連絡しているとフェドロフ副首相	https://www.itmedia.co.jp/news/articles/2203/25/news077.html	reuters	2022-03-25 09:33:00
IT	ITmedia 総合記事一覧	[ITmedia エグゼクティブ] 「人生100年時代」はスマートモビリティ　大阪南部で実証実験　万博での活用も	https://mag.executive.itmedia.co.jp/executive/articles/2203/25/news078.html	公共交通機関	2022-03-25 09:10:00
TECH	Techable（テッカブル）	拡張はこれ一つで！ ストレージもサポートする16ポート搭載のドック「DOCKMULE」	https://techable.jp/archives/175726	dockmule	2022-03-25 00:00:29
AWS	AWS Architecture Blog	Insights for CTOs: Part 3 – Growing your business with modern data capabilities	https://aws.amazon.com/blogs/architecture/insights-for-ctos-part-3-growing-your-business-with-modern-data-capabilities/	Insights for CTOs Part Growing your business with modern data capabilitiesThis post was co wrtiten with Jonathan Hwang head of Foundation Data Analytics at Zendesk In my role as a Senior Solutions Architect I have spoken to chief technology officers CTOs and executive leadership of large enterprises like big banks software as a service SaaS businesses mid sized enterprises and startups In this part series I share …	2022-03-25 00:11:15
デザイン	コリス	デザインが赤ペンで見違える！ 豊富な作例を上司がフィードバック、ブラッシュアップされていく工程がよく分かるデザイン書 -選ばれるデザイナーへの道	https://coliss.com/articles/book-review/isbn-9784802613606.html	続きを読む	2022-03-25 00:36:43
デザイン	コリス	今週末まで有料フォントが無料！ 文字のデザインが楽しめる、ナチュラルな手書き感がかっこいいフォント -Buttergone	https://coliss.com/articles/products/buttergone-by-luckytype.html	続きを読む	2022-03-25 00:26:04
python	Pythonタグが付けられた新着投稿 - Qiita	特定文字で分割して列作成	https://qiita.com/aimainn/items/3f610e491814f1e35e2e		2022-03-25 09:11:38
js	JavaScriptタグが付けられた新着投稿 - Qiita	Splideをスライダーの新たな定番としたく	https://qiita.com/heeroo_ymsw/items/8c2e9cda9ceb884a15a4	どちらも同じような機能を備えていますし、ayも考慮されているため、あとは軽量さと相対単位が使えるという点が魅力なのでSplideを使用します。	2022-03-25 09:34:22
Docker	dockerタグが付けられた新着投稿 - Qiita	docker-compose　volumes定義とは？	https://qiita.com/keitean/items/23c75f834cc839ca9634	また、dockersysteminfoでデフォルトで使用されているvolumeの確認もできます。	2022-03-25 09:18:29
技術ブログ	Developers.IO	【4/26（火）リモート】クラスメソッドの会社説明会を開催します	https://dev.classmethod.jp/news/jobfair-220426/	会社説明会	2022-03-25 00:50:35
海外TECH	DEV Community	Building your first machine learning model	https://dev.to/mage_ai/building-your-first-machine-learning-model-40lc	Building your first machine learning modelWe ll show you how to build a basic machine learning model For a quick background on AI ML please check out this blog post for an overview In this tutorial we ll use the Titanic dataset to predict which passengers survived the crash The dataset includes information on each passenger the cabin they stayed in their gender and more High level steps for building an AI ML modelData preparationChoose algorithmHyperparameter tuningTrain modelEvaluate performanceDeploy Integrate model SetupGo to Google Colaboratory click the top left “File and click “New notebook Download this file called titanic survival csv In your “New notebook on Google Colaboratory click the folder icon in the top left and then drag the file you just downloaded called “titanic survival csv and drop it into that area Data preparationHere are the steps when preparing data Download and split dataAdd columnsRemove columnsImpute valuesScale valuesEncode valuesSelect features Download and split dataWe need to load the data into memory by downloading it from a website a database data warehouse SaaS tool etc Once we download it we can load it in memory to operate on quickly Before we split the data we ll need to determine which column we want to predict In this tutorial we ll predict which passengers survived from sklearn model selection import train test splitimport pandas as pddf pd read csv content titanic survival csv label feature name Survived X df drop columns label feature name y df label feature name After that we need to split the data into parts for training the AI ML model aka train set and for evaluating the performance of the model aka test set The train set will have of the rows from the original data There are different strategies for splitting the data however a common method is to stratify the data so that there is a representative number of rows in both the train set and test set X train raw X test raw y train y test train test split X y stratify y test size Add columnsThe data you downloaded may not have all the columns you need You may want to add a few more columns by combining existing columns or performing some sort of calculation For example you may want to create a column called “year which extracts the year of a date value from the birthday column df X train raw copy Add a column to determine if the person can votedf can vote df Age apply lambda age if age gt else passengers can vote aka they are or olderdf can vote value counts Cabin letter a cabin can be denoted as B The cabin letter will be B df loc cabin letter df Cabin apply lambda cabin cabin if cabin and type cabin is str else None Remove columnsThere may be columns that you don t think the model should learn from For example the model may not care about specific user IDs or email addresses the email domain might matter In these cases we want to remove these columns from the data By removing these columns we help the model focus on what matters instead of trying to make sense of data that has no impact on the prediction For example a passenger s ID probably has very little impact on whether they survived the sinking of the Titanic df df drop columns Name PassengerId Name and PassengerId is no longer a columndf columns tolist Impute valuesYour data may have missing values in a particular column The AI ML model has a hard time knowing what to do with missing values We can help it by filling in those missing values using some heuristic For example there are a lot of missing values in the “Cabin column For those with no known cabin we ll fill in the value “somewhere out of sight For those with missing age we ll use the median age to fill in those missing values from sklearn impute import SimpleImputerprint f Missing values in Cabin len df df Cabin isna index df loc df Cabin isna Cabin somewhere out of sight df loc df cabin letter isna cabin letter ZZZ print f Missing values in Age len df df Age isna index age imputer SimpleImputer strategy median df loc Age age imputer fit transform df Age print f Missing values in Embarked len df df Embarked isna index df loc df Embarked isna Embarked no idea Scale valuesAdjust the values of number columns to fall within similar ranges so that large numbers such as seconds since epoch don t affect the prediction disproportionately as much as smaller values such as age For example if you have a column that is in seconds and a column that is in days the difference in seconds between today and last week is seconds The difference in days between today and last week is If we don t scale these values then the model will think the column with seconds has a greater distance between numbers than the column with days There are multiple scaling strategies such as standard scaler and normalizer For more information check out this thread from sklearn preprocessing import StandardScalerscaler StandardScaler df loc Age scaler fit transform df Age Encode valuesAI ML algorithms perform mathematical operations using numbers We must convert columns that contain strings into a number representation A common technique is to encode categorical values For example we can convert the value “male to and “female to Note we re going to use one hot encoding to convert these strings into numbers For further explanation why check out this thread from sklearn preprocessing import OneHotEncodercategorical columns Pclass Sex Embarked cabin letter categorical encoder OneHotEncoder handle unknown ignore categorical encoder fit df categorical columns Add the new columns to the datanew column names for idx cat column name in enumerate categorical columns values categorical encoder categories idx new column names f cat column name value for value in values df loc new column names categorical encoder transform df categorical columns toarray Select featuresNow that we ve prepared our data we need to select the features we want our model to learn from There are many techniques for doing this Mage s tool handles this automatically for you For this tutorial we ll simply select the features we manually added scaled or encoded features to use Age SibSp Parch Fare can vote new column namesX train df features to use copy Choose algorithmOnce our data is in a state that is ready to be trained on we must choose an algorithm to use Different algorithms are best suited for different types of problems and different types of data For this tutorial we ll use a basic algorithm called logistic regression that ll help us classify which passengers survived the Titanic crash from sklearn linear model import LogisticRegressionclassifier LogisticRegression max iter Hyperparameter tuningAn AI ML model has parameters that aren t related to the features aka columns in the data These “hyper parameters control how the model behaves throughout its training When improving AI ML models it s common to try a bunch of different combinations of hyperparameters that ll yield the best results We ll skip this optimization for this tutorial keep an eye out for a future article on this topic Train modelWe take the data that was prepared X train and the actual results y train for each row e g whether the passenger survived the Titanic and feed it into the model The model will learn from looking at the values in each column and seeing what result it produces for survived for not survived Once the model learns from all the data it will finish training and can be used to make predictions on unseen data classifier fit X train y train Evaluate performancePrepare test dataUse model to predict on test dataCalculate model accuracyDetermine baseline performance and compare Prepare test dataFirst we ll prepare our test data e g add columns remove columns impute values scale values encode values and select features in the same way we did for our train set One caveat is that we won t “fit our standard scaler or our encoders because we only want to “fit those on the train set Note the code below is an exact copy of the code written above during data preparation for the train set except we are calling functions on the variable containing the test data A better engineering practice would be to refactor the code by creating a reusable function that accepts a Pandas dataframe as an argument calls all the data preparation steps on that dataframe and returns it Here is the code written and not refactored for clarity sake X test X test raw copy Add columnsX test can vote X test Age apply lambda age if age gt else X test loc cabin letter X test Cabin apply lambda cabin cabin if cabin and type cabin is str else None Remove columnsX test X test drop columns Name PassengerId Impute valuesX test loc X test Cabin isna Cabin somewhere out of sight X test loc X test cabin letter isna cabin letter ZZZ X test loc Age age imputer transform X test Age X test loc X test Embarked isna Embarked no idea Scale columnsX test loc Age scaler transform X test Age Encode valuesX test loc new column names categorical encoder transform X test categorical columns toarray Select featuresX test X test features to use copy Use model to predict on test dataNext we use the model to predict who survives from the test data remember we split the data earlier during data preparation y pred classifier predict X test Calculate model accuracyRegression and classification models have different metrics that are used to evaluate the performance of the model Since we re using a classification model even though it s called logistic regression it can be used for classifying we ll use accuracy as our metric If there were multiple categories we re predicting we ll also want to use precision and recall as a metric from sklearn metrics import accuracy scoreaccuracy accuracy score y test y pred print f Accuracy score accuracy Determine baseline performance and compareIn order for us to understand how good this accuracy is we need to establish a baseline In this specific example the baseline accuracy will be the number of people who didn t survive within the test set divided by the number of rows in the test set baseline accuracy score y test value counts len y test print f Model performance accuracy print f Baseline performance baseline accuracy score Deploy Integrate modelOnce you trained the model and fine tuned it to your business needs it s time to integrate it into your product or business operations There are several ways of doing this you can deploy the model to an online server where the model s prediction can be accessed via an API request or you can set up your model to perform batch predictions and export those predictions to your data warehouse data lake etc Deploying your model maintaining the model keeping it up to date so that it makes relevant predictions and making sure your feature data is fresh and readily available to retrieve for online predictions is time consuming costly extremely complex and a non differentiating skillset Instead of focusing your energy on this particular aspect it s common to rely on other tools for this service A tool like Mage not only helps you prepare your data and train your model it also helps you access your model from an API endpoint and keeps the model relevant by retraining it regularly ConclusionEasy AI ML development tutorialHere is the link to the entire code Additional resourcesBest book for breaking into machine learning Hands On Machine Learning with Scikit Learn Keras and TensorFlow Concepts Tools and Techniques to Build Intelligent Systems	2022-03-25 00:40:23
海外TECH	DEV Community	March 24th, 2022: What did you learn this week?	https://dev.to/nickytonline/march-24th-2022-what-did-you-learn-this-week-589	March th What did you learn this week It s that time of the week again So wonderful devs what did you learn this week It could be programming tips career advice etc Feel free to comment with what you learnt and or reference your TIL post to give it some more exposure ltag tag id follow action button background color ffedc important color important border color ffedc important todayilearned Follow Summarize a concept that is new to you	2022-03-25 00:38:25
Apple	AppleInsider - Frontpage News	New 15-inch MacBook may not be MacBook Air, says Kuo	https://appleinsider.com/articles/22/03/25/new-15-inch-macbook-may-not-be-macbook-air-says-kuo?utm_medium=rss	New inch MacBook may not be MacBook Air says KuoMing Chi Kuo predicts Apple will add a new inch MacBook to its lineup but believes that the model may not be another MacBook Air and it won t arrive for well over a year A report on Wednesday claimed that Apple may be bringing a inch MacBook to market in with analysts assuming that the model would fit into Apple s MacBook Air lineup However Apple Analyst Ming Chi Kuo took to Twitter on Thursday to refute at least some of that report Read more	2022-03-25 00:29:40
海外TECH	Engadget	Ex-TikTok moderators sue over 'emotional distress' from disturbing videos	https://www.engadget.com/ex-tik-tok-moderators-sue-over-emotional-distress-from-disturbing-videos-000736797.html?src=rss	Ex TikTok moderators sue over x emotional distress x from disturbing videosTwo former TikTok moderators filed a federal lawsuit seeking class action status today against the platform and parent company Bytedance reportedNPR The plaintiffs Ashley Velez and Reece Young worked for the social video platform last year as contractors To fulfill their role as moderators they witnessed “many acts of extreme and graphic violence including murder bestiality necrophilia and other disturbing images The lawsuit accuses TikTok of negligence and violating labor laws in California the state where the platform s US operations is based Both plaintiffs said they were tasked with viewing hours of disturbing footage often working hour days They both paid for counseling out of pocket in order to deal with the psychological toll of the job The lawsuit accuses TikTok of imposing high “productivity standards on moderators which forced them to watch large volumes of disturbing content without a break Both employees were also forced to sign non disclosure agreements as a condition of their employment quot We would see death and graphic graphic pornography I would see nude underage children every day quot Velez told NPR quot I would see people get shot in the face and another video of a kid getting beaten made me cry for two hours straight quot Moderators at Facebook and other platforms have spoken out in the past about the severe psychological toll of their jobs Employees have alleged they re given a short period of time usually only seconds to determine whether a video violates the platform s policies The job has often been called “the worst job in technology quot and workers regularly suffer from depression PTSD like symptoms and suicidal ideation In a settlement Facebook paid over million to a group of former moderators who said they developed PTSD from the job This is not the first lawsuit of this type for TikTok which currently has a base of content moderators worldwide Last December another content moderator for TikTok also sued the platform for negligence and violating workplace safety standards According to NPR the lawsuit was dropped last month after the plaintiff was fired	2022-03-25 00:07:36
海外科学	NYT > Science	British Rainfall Records Extended Back to 1836 Thanks to Covid Lockdowns	https://www.nytimes.com/2022/03/24/climate/britain-rainfall-historical-data.html	British Rainfall Records Extended Back to Thanks to Covid LockdownsAs Britain went into its first Covid lockdown a scientist asked for help transcribing rainfall records spanning three centuries Thousands of people online answered the call	2022-03-25 00:47:01
海外科学	NYT > Science	Methane Leaks Plague New Mexico Oil and Gas Wells	https://www.nytimes.com/2022/03/24/climate/methane-leaks-new-mexico.html	Methane Leaks Plague New Mexico Oil and Gas WellsAn analysis found leaks of methane a potent greenhouse gas from oil and gas drilling in the Permian Basin were many times higher than government estimates	2022-03-25 00:50:43
海外TECH	WIRED	Feds Allege Destructive Russian Hackers Targeted US Refineries	https://www.wired.com/story/triton-berserk-bear-russian-hackers-doj-indictment	Feds Allege Destructive Russian Hackers Targeted US RefineriesThe Justice Department unsealed indictments against four alleged Russian hackers said to have targeted US energy infrastructure for nearly a decade	2022-03-25 00:21:15
金融	ニッセイ基礎研究所	コロナ禍における高齢者の移動の減少と健康悪化への懸念～先行研究のレビューとニッセイ基礎研究所のコロナ調査から～	https://www.nli-research.co.jp/topics_detail1/id=70647?site=nli	目次ーはじめにーコロナ禍による高齢者の移動の減少コロナ禍による日常生活の変化高齢者の移動の減少と健康への影響ー高齢者の外出と心身の健康との関連先行研究のレビューより歩行による死亡率低下や認知症予防への効果社会参加による死亡率低下や認知症予防への効果外出頻度の重要性小括ー終わりに高齢者の健康維持を目的とした「新しい生活様式」構築を言新型コロナウイルス感染拡大後、不要不急の外出自粛や在宅勤務の推奨、飲食店の営業時間短縮などにより、人々の移動の機会が大きく減った。	2022-03-25 09:26:37
ニュース	BBC News - Home	Ukraine War: Civilians abducted as Russia tries to assert control	https://www.bbc.co.uk/news/world-europe-60858363?at_medium=RSS&at_campaign=KARANGA	journalists	2022-03-25 00:02:58
ニュース	BBC News - Home	Diary giant warns of supply issues unless wages rise	https://www.bbc.co.uk/news/business-60825516?at_medium=RSS&at_campaign=KARANGA	prices	2022-03-25 00:01:26
ニュース	BBC News - Home	Mobile loophole for gaming drivers is closed	https://www.bbc.co.uk/news/uk-60869941?at_medium=RSS&at_campaign=KARANGA	britain	2022-03-25 00:53:09
ニュース	BBC News - Home	Ukraine war: Return to Irpin, the town destroyed in the battle for Kyiv	https://www.bbc.co.uk/news/world-europe-60867789?at_medium=RSS&at_campaign=KARANGA	ukrainian	2022-03-25 00:53:01
ニュース	BBC News - Home	Notts County and the conman: Following your team through a football scam	https://www.bbc.co.uk/sport/football/60794963?at_medium=RSS&at_campaign=KARANGA	Notts County and the conman Following your team through a football scamIn the summer of mysterious new owners promised to propel fourth tier side Notts County to grand success It was all a big lie	2022-03-25 00:01:03
GCP	Google Cloud Platform Japan 公式ブログ	浜松市役所：Chromebook が行政サービスの DX 推進の要となり、高いセキュリティと機動力の両立を実現	https://cloud.google.com/blog/ja/topics/customers/hamamatsu-city-chromebooks-drive-dx/	」成瀬氏これまで部署ごとで数台のインターネット環境を共有するという運用を行って来ましたが、コロナ禍によってインターネット上での業務が急増し、より多くのインターネットに接続できる環境が求められていることも追い風になっていると言います。	2022-03-25 02:00:00
北海道	北海道新聞	米大統領、ロシアのＧ２０排除を　化学兵器に「相応の対応取る」	https://www.hokkaido-np.co.jp/article/660944/	化学兵器	2022-03-25 09:30:00
北海道	北海道新聞	雪さらしで作るジーンズ、新潟　妙高の服飾店、伝統技法活用	https://www.hokkaido-np.co.jp/article/660924/	新潟妙高	2022-03-25 09:12:30
北海道	北海道新聞	反差別集めた大阪の副読本の展示　「にんげん」の軌跡たどる	https://www.hokkaido-np.co.jp/article/660923/	大阪府内	2022-03-25 09:12:30
北海道	北海道新聞	仏大統領、遠回しに中国けん制　「ロシアを支援しないと信じる」	https://www.hokkaido-np.co.jp/article/660943/	遠回し	2022-03-25 09:26:00
北海道	北海道新聞	「Ｗ杯でベスト８以上を」　サッカー森保監督、一夜明け	https://www.hokkaido-np.co.jp/article/660926/	監督	2022-03-25 09:26:13
北海道	北海道新聞	大リーグ、筒香がオープン戦１号　大谷は２四球、鈴木は出場せず	https://www.hokkaido-np.co.jp/article/660942/	大リーグ	2022-03-25 09:14:00
北海道	北海道新聞	食用魚介類の自給率９４％を目標　水産計画、不漁対策を強化	https://www.hokkaido-np.co.jp/article/660936/	基本計画	2022-03-25 09:07:00
北海道	北海道新聞	成年後見人の交代柔軟に　必要時だけ利用、計画決定	https://www.hokkaido-np.co.jp/article/660935/	成年後見人	2022-03-25 09:22:18
マーケティング	MarkeZine	数字が見えるとおもしろさが一気に増していく。BtoBマーケターの成長と仕事の魅力	http://markezine.jp/article/detail/38592	BtoBマーケターの成長と仕事の魅力BtoBマーケティング組織の立ち上げから短期間で成果を残し、賞を受けるなど注目を集めたブラザー販売。	2022-03-25 09:30:00
ニュース	THE BRIDGE	法律トラブル解決支援プラットフォーム開発のキビタス、8,000万円を資金調達——XTech V、ANOBAKA、クオンタムリープVから	https://thebridge.jp/2022/03/civitas-jpy80m-funding	法律トラブル解決支援プラットフォーム開発のキビタス、万円を資金調達ーXTechV、ANOBAKA、クオンタムリープVから法律トラブル解決支援ODROnlineDisputeResolutionプラットフォームを開発するキビタスは日、直近のラウンドで万円を調達したと発表した。	2022-03-25 00:00:54
GCP	Cloud Blog JA	浜松市役所：Chromebook が行政サービスの DX 推進の要となり、高いセキュリティと機動力の両立を実現	https://cloud.google.com/blog/ja/topics/customers/hamamatsu-city-chromebooks-drive-dx/	」成瀬氏これまで部署ごとで数台のインターネット環境を共有するという運用を行って来ましたが、コロナ禍によってインターネット上での業務が急増し、より多くのインターネットに接続できる環境が求められていることも追い風になっていると言います。	2022-03-25 02:00:00