Les bookmarks de Clem
2385 shaares
Idem
Super présentation, avec plein de sources
Très intéressant !
De toute façon sans s pour dire quoiqu'il en soit.
Super source de thèmes pour la préparation de présentation
Super conseils pour préparer des présentations
Une réponse super-claire à une question que je me posais depuis longtemps. Quelle erreur max peut-on attendre sur une mesure de distance si l'on considère la terre comme une sphère. Réponse, assez peu étonnante : 22km (ce qui correspond à la différence entre le rayon à l'équateur et celui au pôle). Le gars fait une superbe carto des erreurs de distance à partir d'un point localisé en Camargue, et on peut voir que l'erreur monte vite. De la Camargue en Russie, ça peut faire déjà 5km de différence. Mais, en termes d'erreur relative, ça ne dépassera jamais 0.5% de la distance réelle.
Marrant : ils prennent une image en bitmap (noir et blanc), échantillonnent des pixels, et utilisent un algo pour trouver le plus court chemin passant par les pixels échantillonnés (pb du voyageur de commerce). Résultat: un portrait en une ligne.
Le mec va plus loin: il va prendre une image en niveaux de gris, et va échantillonner les pixels proportionnellement au poids défini par le niveau de gris. Il répète le processus 250 fois, et ça rend l'impression d'une image faite au crayon. Ici : https://fronkonstin.com/2018/04/17/pencil-scribbles/
Le mec va plus loin: il va prendre une image en niveaux de gris, et va échantillonner les pixels proportionnellement au poids défini par le niveau de gris. Il répète le processus 250 fois, et ça rend l'impression d'une image faite au crayon. Ici : https://fronkonstin.com/2018/04/17/pencil-scribbles/
L'estimation d'erreurs types pour les régressions lasso, c'est visiblement un sacré bordel.
Super explication de pourquoi on se fout de la surdispersion en régression lasso.
La vignette est sympa
De l'usage responsable des drones pour filmer la faune.
Semble assez utile
Sur l'introversion en science.
Intéressant: il utilise la loi gamma pour simuler une quasi-poisson. Malheureusement, ses liens ne marchent pas. Mais je trouve l'idée intéressante, je me la garde sous le coude.
Connaissais pas.
Super tutos sur R, sf et ggplot2, de Mel Moreno et Mathieu Basille. Transféré à tous les collègues.
Oh mais il y a des choses ici !
Intéressant, sur la difficulté de communiquer de façon impartiale en journalisme et science. Extraits choisis. Le gars bossait dans la section communication scientifique à la BBC, et a commencé à remettre en question sa façon de bosser en bossant au Winston centre:
"Officially we were supposed to ‘inform, educate, entertain’ and we thought we knew how to do all three. Now, though, I am beginning to question what exactly science communication is doing.
(...) Evidence presented purely to inform — balanced and not trying to persuade a person to agree or disagree — does not make for entertainment, almost by definition.
(...) in order to make a good decision, researchers believe that we first need to make ourselves imagine more than one potential future scenario. We need to open our minds to the possibility that things could turn out well or badly. We need emotion.
(...) We all have a right to be ‘informed and not persuaded’ — an ethical right and often a legal right. And yet training in communication focuses almost entirely on how to grab attention and how to manipulate emotions to tell a story — how to be persuasive. There is very little training on how to use emotions more subtly, in a way that opens minds to possibilities constructively, but is not designed to persuade (I’m certainly still learning how it might be done).
Of course in a competitive environment it is important to get your voice heard, and I often hear people say their objective is to persuade people to ‘do the right thing’. But who is defining ‘right’? If you as the communicator are — if you are defining the story — then you are not simply informing.
When people are making really important decisions — decisions about their health, their finances, about policies that will affect millions, or about someone’s guilt or innocence — I would now argue that a different kind of communication is needed: the skill to engage and be clear and to allow the audience to form their own story through the information, and to make their own decision at the end as to how to react to it."
"Officially we were supposed to ‘inform, educate, entertain’ and we thought we knew how to do all three. Now, though, I am beginning to question what exactly science communication is doing.
(...) Evidence presented purely to inform — balanced and not trying to persuade a person to agree or disagree — does not make for entertainment, almost by definition.
(...) in order to make a good decision, researchers believe that we first need to make ourselves imagine more than one potential future scenario. We need to open our minds to the possibility that things could turn out well or badly. We need emotion.
(...) We all have a right to be ‘informed and not persuaded’ — an ethical right and often a legal right. And yet training in communication focuses almost entirely on how to grab attention and how to manipulate emotions to tell a story — how to be persuasive. There is very little training on how to use emotions more subtly, in a way that opens minds to possibilities constructively, but is not designed to persuade (I’m certainly still learning how it might be done).
Of course in a competitive environment it is important to get your voice heard, and I often hear people say their objective is to persuade people to ‘do the right thing’. But who is defining ‘right’? If you as the communicator are — if you are defining the story — then you are not simply informing.
When people are making really important decisions — decisions about their health, their finances, about policies that will affect millions, or about someone’s guilt or innocence — I would now argue that a different kind of communication is needed: the skill to engage and be clear and to allow the audience to form their own story through the information, and to make their own decision at the end as to how to react to it."