@techandstream.bsky.social
Pinned
Fonction d’activation neuronale
Softmax softmax(xᵢ) = eˣⁱ / ∑eˣʲ Used for multi-class classification. Outputs probability distribution.
Softmax softmax(xᵢ) = eˣⁱ / ∑eˣʲ Used for multi-class classification. Outputs probability distribution.
Fonction d’activation neuronale
Softmax softmax(xᵢ) = eˣⁱ / ∑eˣʲ Used for multi-class classification. Outputs probability distribution.
kvnbbg.fr
November 11, 2025 at 6:57 PM
Fonction d’activation neuronale
Softmax softmax(xᵢ) = eˣⁱ / ∑eˣʲ Used for multi-class classification. Outputs probability distribution.
Softmax softmax(xᵢ) = eˣⁱ / ∑eˣʲ Used for multi-class classification. Outputs probability distribution.
Fonction d’activation neuronale
ReLU ReLU(x) = max(0, x) Most popular. Computationally efficient, helps with vanishing gradient problem.
ReLU ReLU(x) = max(0, x) Most popular. Computationally efficient, helps with vanishing gradient problem.
Fonction d’activation neuronale
ReLU ReLU(x) = max(0, x) Most popular. Computationally efficient, helps with vanishing gradient problem.
kvnbbg.fr
November 11, 2025 at 6:56 PM
Fonction d’activation neuronale
ReLU ReLU(x) = max(0, x) Most popular. Computationally efficient, helps with vanishing gradient problem.
ReLU ReLU(x) = max(0, x) Most popular. Computationally efficient, helps with vanishing gradient problem.
Fonction d’activation neuronale
Tanh tanh(x) = (eˣ - e⁻ˣ) / (eˣ + e⁻ˣ) Output range: (-1, 1). Zero-centered, often better than sigmoid.
Tanh tanh(x) = (eˣ - e⁻ˣ) / (eˣ + e⁻ˣ) Output range: (-1, 1). Zero-centered, often better than sigmoid.
Fonction d’activation neuronale
Tanh tanh(x) = (eˣ - e⁻ˣ) / (eˣ + e⁻ˣ) Output range: (-1, 1). Zero-centered, often better than sigmoid.
kvnbbg.fr
November 11, 2025 at 6:54 PM
Fonction d’activation neuronale
Tanh tanh(x) = (eˣ - e⁻ˣ) / (eˣ + e⁻ˣ) Output range: (-1, 1). Zero-centered, often better than sigmoid.
Tanh tanh(x) = (eˣ - e⁻ˣ) / (eˣ + e⁻ˣ) Output range: (-1, 1). Zero-centered, often better than sigmoid.
Fonction d’activation neuronale
Sigmoid σ(x) = 1 / (1 + e⁻ˣ) Output range: (0, 1). Used for binary classification and probability outputs
Sigmoid σ(x) = 1 / (1 + e⁻ˣ) Output range: (0, 1). Used for binary classification and probability outputs
Fonction d’activation neuronale
Sigmoid σ(x) = 1 / (1 + e⁻ˣ) Output range: (0, 1). Used for binary classification and probability outputs
kvnbbg.fr
November 11, 2025 at 6:53 PM
Fonction d’activation neuronale
Sigmoid σ(x) = 1 / (1 + e⁻ˣ) Output range: (0, 1). Used for binary classification and probability outputs
Sigmoid σ(x) = 1 / (1 + e⁻ˣ) Output range: (0, 1). Used for binary classification and probability outputs
Type System: statique, dynamique ou sadique ?
en.wikipedia.org/wiki/Type_system Le typage dépendant (dans Epigram, Idris, Agda, Dependent ML) se distingue du typage statique classique (comme Java, C, Rust) et du typage dynamique (comme Python, Ruby, JavaScript) parce qu’il ne se contente pas de…
en.wikipedia.org/wiki/Type_system Le typage dépendant (dans Epigram, Idris, Agda, Dependent ML) se distingue du typage statique classique (comme Java, C, Rust) et du typage dynamique (comme Python, Ruby, JavaScript) parce qu’il ne se contente pas de…
Type System: statique, dynamique ou sadique ?
en.wikipedia.org/wiki/Type_system Le typage dépendant (dans Epigram, Idris, Agda, Dependent ML) se distingue du typage statique classique (comme Java, C, Rust) et du typage dynamique (comme Python, Ruby, JavaScript) parce qu’il ne se contente pas de vérifier que « les types correspondent » ou d’attendre l’exécution pour signaler une erreur, mais il encode dans les types eux-mêmes des propriétés logiques du programme (par exemple Vect n a pour exprimer la longueur dans le type, append : Vect n a -> Vect m a -> Vect (n + m) a pour garantir la taille du résultat, ou safeHead : Vect (S n) a -> a pour rendre impossible l’accès à une liste vide), de sorte que le compilateur prouve la cohérence du programme avant qu’il s’exécute, reflétant une philosophie développeur où l’on ne « teste pas pour voir si ça marche », mais où l’on écrit du code qui ne peut pas ne pas être correct, faisant de la programmation non pas un pari sur l’exécution, mais une construction explicite de certitude.
kvnbbg.fr
November 11, 2025 at 3:08 PM
Type System: statique, dynamique ou sadique ?
en.wikipedia.org/wiki/Type_system Le typage dépendant (dans Epigram, Idris, Agda, Dependent ML) se distingue du typage statique classique (comme Java, C, Rust) et du typage dynamique (comme Python, Ruby, JavaScript) parce qu’il ne se contente pas de…
en.wikipedia.org/wiki/Type_system Le typage dépendant (dans Epigram, Idris, Agda, Dependent ML) se distingue du typage statique classique (comme Java, C, Rust) et du typage dynamique (comme Python, Ruby, JavaScript) parce qu’il ne se contente pas de…
Le français qui a update nos base de données, Pierre-Simon Laplace. La croyance mise à jour s’appelle posterior
the guy who developed Bayesian stats as it is today is a French mathematician called Pierre-Simon Laplace on peut se référer aux travaux de Pierre-Simon Laplace tels qu’ils sont…
the guy who developed Bayesian stats as it is today is a French mathematician called Pierre-Simon Laplace on peut se référer aux travaux de Pierre-Simon Laplace tels qu’ils sont…
Le français qui a update nos base de données, Pierre-Simon Laplace. La croyance mise à jour s’appelle posterior
the guy who developed Bayesian stats as it is today is a French mathematician called Pierre-Simon Laplace on peut se référer aux travaux de Pierre-Simon Laplace tels qu’ils sont présentés dans des ouvrages comme « Théorie analytique des probabilités » ou encore dans des articles récents de revues de statistiques, comme le « Journal of Statistical Physics ». prior = Belief(value=0.3)
kvnbbg.fr
November 11, 2025 at 8:18 AM
Le français qui a update nos base de données, Pierre-Simon Laplace. La croyance mise à jour s’appelle posterior
the guy who developed Bayesian stats as it is today is a French mathematician called Pierre-Simon Laplace on peut se référer aux travaux de Pierre-Simon Laplace tels qu’ils sont…
the guy who developed Bayesian stats as it is today is a French mathematician called Pierre-Simon Laplace on peut se référer aux travaux de Pierre-Simon Laplace tels qu’ils sont…
Evidance changes your beliefs
Extraordinary beliefs. require extraordinary evidence. And so that's basically what Bayes' formula is about. It's the mathematical way of saying extraordinary claims require extraordinary evidence. And so if you think I'm an alien, a priori, it's a very small…
Extraordinary beliefs. require extraordinary evidence. And so that's basically what Bayes' formula is about. It's the mathematical way of saying extraordinary claims require extraordinary evidence. And so if you think I'm an alien, a priori, it's a very small…
Evidance changes your beliefs
Extraordinary beliefs. require extraordinary evidence. And so that's basically what Bayes' formula is about. It's the mathematical way of saying extraordinary claims require extraordinary evidence. And so if you think I'm an alien, a priori, it's a very small probability. But then let's say you get access to one of my blood tests. And in this resource, you get the information that actually my blood is blue and not red like a human.
kvnbbg.fr
November 11, 2025 at 8:01 AM
Evidance changes your beliefs
Extraordinary beliefs. require extraordinary evidence. And so that's basically what Bayes' formula is about. It's the mathematical way of saying extraordinary claims require extraordinary evidence. And so if you think I'm an alien, a priori, it's a very small…
Extraordinary beliefs. require extraordinary evidence. And so that's basically what Bayes' formula is about. It's the mathematical way of saying extraordinary claims require extraordinary evidence. And so if you think I'm an alien, a priori, it's a very small…
What is Bayesian mindset ? a constant process of updates
It's extremely painful to have a perfectionist mindset. It's a way to control the world around you. What helps me is to compare myself to what I was yesterday. Because the problem is that we try to compare ourselves with someone who's maybe…
It's extremely painful to have a perfectionist mindset. It's a way to control the world around you. What helps me is to compare myself to what I was yesterday. Because the problem is that we try to compare ourselves with someone who's maybe…
What is Bayesian mindset ? a constant process of updates
It's extremely painful to have a perfectionist mindset. It's a way to control the world around you. What helps me is to compare myself to what I was yesterday. Because the problem is that we try to compare ourselves with someone who's maybe spent hours mastering a skill that we just started with. Or comparing yourself with your ideal version of yourself.
kvnbbg.fr
November 11, 2025 at 7:50 AM
What is Bayesian mindset ? a constant process of updates
It's extremely painful to have a perfectionist mindset. It's a way to control the world around you. What helps me is to compare myself to what I was yesterday. Because the problem is that we try to compare ourselves with someone who's maybe…
It's extremely painful to have a perfectionist mindset. It's a way to control the world around you. What helps me is to compare myself to what I was yesterday. Because the problem is that we try to compare ourselves with someone who's maybe…
Entre histoire et data: penser contre soi-même, ça forge
« Who writes history books and who analyzes data and comes up with the statistics in a way determines the way people get this information? » Le mec qui a étudié l’histoire, même s’il fait maintenant des stats en Python, il reste influencé…
« Who writes history books and who analyzes data and comes up with the statistics in a way determines the way people get this information? » Le mec qui a étudié l’histoire, même s’il fait maintenant des stats en Python, il reste influencé…
Entre histoire et data: penser contre soi-même, ça forge
« Who writes history books and who analyzes data and comes up with the statistics in a way determines the way people get this information? » Le mec qui a étudié l’histoire, même s’il fait maintenant des stats en Python, il reste influencé par la narration, la structure, la chronologie. Et inversement, un pur data scientist peut finir par écrire une histoire à travers ses graphes
kvnbbg.fr
November 11, 2025 at 7:14 AM
Entre histoire et data: penser contre soi-même, ça forge
« Who writes history books and who analyzes data and comes up with the statistics in a way determines the way people get this information? » Le mec qui a étudié l’histoire, même s’il fait maintenant des stats en Python, il reste influencé…
« Who writes history books and who analyzes data and comes up with the statistics in a way determines the way people get this information? » Le mec qui a étudié l’histoire, même s’il fait maintenant des stats en Python, il reste influencé…
Mélancolie, poésie et humeur
C’est vrai que l’ambiance d’une ville peut vraiment influencer notre humeur. Chaque endroit a un peu sa propre atmosphère, et parfois on ressent ça sans même s’en rendre compte. Donc c’est tout à fait possible que l’énergie de l’endroit où tu vis joue un rôle sur ce…
C’est vrai que l’ambiance d’une ville peut vraiment influencer notre humeur. Chaque endroit a un peu sa propre atmosphère, et parfois on ressent ça sans même s’en rendre compte. Donc c’est tout à fait possible que l’énergie de l’endroit où tu vis joue un rôle sur ce…
Mélancolie, poésie et humeur
C’est vrai que l’ambiance d’une ville peut vraiment influencer notre humeur. Chaque endroit a un peu sa propre atmosphère, et parfois on ressent ça sans même s’en rendre compte. Donc c’est tout à fait possible que l’énergie de l’endroit où tu vis joue un rôle sur ce que tu ressens
kvnbbg.fr
November 11, 2025 at 7:02 AM
Mélancolie, poésie et humeur
C’est vrai que l’ambiance d’une ville peut vraiment influencer notre humeur. Chaque endroit a un peu sa propre atmosphère, et parfois on ressent ça sans même s’en rendre compte. Donc c’est tout à fait possible que l’énergie de l’endroit où tu vis joue un rôle sur ce…
C’est vrai que l’ambiance d’une ville peut vraiment influencer notre humeur. Chaque endroit a un peu sa propre atmosphère, et parfois on ressent ça sans même s’en rendre compte. Donc c’est tout à fait possible que l’énergie de l’endroit où tu vis joue un rôle sur ce…
Creates an array [1, 2, 3, 4, 5]
A function pointer is a variable that stores the memory address of a function. This allows us to treat code as data, we can pass functions as arguments and decide which function to call at runtime.
A function pointer is a variable that stores the memory address of a function. This allows us to treat code as data, we can pass functions as arguments and decide which function to call at runtime.
Creates an array [1, 2, 3, 4, 5]
A function pointer is a variable that stores the memory address of a function. This allows us to treat code as data, we can pass functions as arguments and decide which function to call at runtime.
kvnbbg.fr
November 10, 2025 at 3:31 AM
Creates an array [1, 2, 3, 4, 5]
A function pointer is a variable that stores the memory address of a function. This allows us to treat code as data, we can pass functions as arguments and decide which function to call at runtime.
A function pointer is a variable that stores the memory address of a function. This allows us to treat code as data, we can pass functions as arguments and decide which function to call at runtime.
Reposted
Tag yourself; I'm "tonight's flight into Atlanta for #KubeCon is currently delayed 3.5 hours and who knows if I should even go to the airport lolsob". www.flightaware.com/miserymap/
November 8, 2025 at 10:39 PM
Tag yourself; I'm "tonight's flight into Atlanta for #KubeCon is currently delayed 3.5 hours and who knows if I should even go to the airport lolsob". www.flightaware.com/miserymap/
Reposted
Reposted
Yeah so they’re cutting flights at the 40 busiest airports starting with 4% on Friday and ramping up to 10% through the weekend, and that includes ATL. If you’re attending #KubeCon and you aren’t threat modeling around this and making contingency plans already, strongly recommend starting now
So, uh, KubeCon attendees: how we feeling about the threat model of them shutting down the airspace? Are we all moving to Atlanta
November 6, 2025 at 3:32 AM
Yeah so they’re cutting flights at the 40 busiest airports starting with 4% on Friday and ramping up to 10% through the weekend, and that includes ATL. If you’re attending #KubeCon and you aren’t threat modeling around this and making contingency plans already, strongly recommend starting now
Reposted
Want the official pass to the coolest #KubeCon party? Register NOW for Southern Sips & Source. Meet the TNS team and leaders from AWS & NVIDIA for a panel on AI Deployment and K8s Optimization.
Save your spot ➡️ airtable.com/apprFEg0UX01...
Save your spot ➡️ airtable.com/apprFEg0UX01...
November 9, 2025 at 11:11 PM
Want the official pass to the coolest #KubeCon party? Register NOW for Southern Sips & Source. Meet the TNS team and leaders from AWS & NVIDIA for a panel on AI Deployment and K8s Optimization.
Save your spot ➡️ airtable.com/apprFEg0UX01...
Save your spot ➡️ airtable.com/apprFEg0UX01...
Reposted
Admin/user/package/changelog docs here: synapse.docs.vertex.link/projects/rap...
synapse-censys — Synapse Power-Ups documentation
synapse.docs.vertex.link
October 15, 2025 at 7:43 PM
Admin/user/package/changelog docs here: synapse.docs.vertex.link/projects/rap...
Reposted
We’re excited to announce that The Vertex Project ( @vertexproject.bsky.social ) is a sponsor of State of Statecraft 2025. Their generous support helps make our event possible.
The Vertex Project creates technology for analytical teams to provide intelligence-driven insights to decision makers.
The Vertex Project creates technology for analytical teams to provide intelligence-driven insights to decision makers.
October 21, 2025 at 10:20 AM
We’re excited to announce that The Vertex Project ( @vertexproject.bsky.social ) is a sponsor of State of Statecraft 2025. Their generous support helps make our event possible.
The Vertex Project creates technology for analytical teams to provide intelligence-driven insights to decision makers.
The Vertex Project creates technology for analytical teams to provide intelligence-driven insights to decision makers.
Reposted
As espionage & criminal lines vanish, precision matters.
Synapse’s modeling flexibility shows who did what, how they’re linked, and where access is reused.
Want to see for yourself and take Synapse for a spin? Request a demo instance here: vertex.link/request-a-demo
Synapse’s modeling flexibility shows who did what, how they’re linked, and where access is reused.
Want to see for yourself and take Synapse for a spin? Request a demo instance here: vertex.link/request-a-demo
The Vertex Project
vertex.link
October 24, 2025 at 7:25 PM
As espionage & criminal lines vanish, precision matters.
Synapse’s modeling flexibility shows who did what, how they’re linked, and where access is reused.
Want to see for yourself and take Synapse for a spin? Request a demo instance here: vertex.link/request-a-demo
Synapse’s modeling flexibility shows who did what, how they’re linked, and where access is reused.
Want to see for yourself and take Synapse for a spin? Request a demo instance here: vertex.link/request-a-demo
Reposted
Synapse’s data model captures inter-group collaboration with:
🔹 Entity:relationship forms (e.g. acts-as-initial-access-broker, develops-tools-for)
🔹 ou:campaign to represent multi-actor ops while preserving per-actor subsets
🔹 risk:attack for granular step-level attribution
🔹 Entity:relationship forms (e.g. acts-as-initial-access-broker, develops-tools-for)
🔹 ou:campaign to represent multi-actor ops while preserving per-actor subsets
🔹 risk:attack for granular step-level attribution
October 24, 2025 at 7:25 PM
Synapse’s data model captures inter-group collaboration with:
🔹 Entity:relationship forms (e.g. acts-as-initial-access-broker, develops-tools-for)
🔹 ou:campaign to represent multi-actor ops while preserving per-actor subsets
🔹 risk:attack for granular step-level attribution
🔹 Entity:relationship forms (e.g. acts-as-initial-access-broker, develops-tools-for)
🔹 ou:campaign to represent multi-actor ops while preserving per-actor subsets
🔹 risk:attack for granular step-level attribution
Reposted
Researchers are seeing growing collaboration between distinct threat actors including access brokers, tool developers, and espionage groups working together.
What looks like one actor might actually be several.
What looks like one actor might actually be several.
October 24, 2025 at 7:25 PM
Researchers are seeing growing collaboration between distinct threat actors including access brokers, tool developers, and espionage groups working together.
What looks like one actor might actually be several.
What looks like one actor might actually be several.
Reposted
Join us the day before CYBERWARCON (Nov 18) for a hands-on Synapse challenge! Investigate a real-world data theft scenario, eat snacks, & earn bragging rights + our first-ever custom challenge coin. vertex.link/events/cyber...
November 3, 2025 at 3:06 PM
Join us the day before CYBERWARCON (Nov 18) for a hands-on Synapse challenge! Investigate a real-world data theft scenario, eat snacks, & earn bragging rights + our first-ever custom challenge coin. vertex.link/events/cyber...
Reposted
Ready to put your analysis skills to the test? Join us on Nov 18 (pre-CYBERWARCON) for a Synapse challenge using a real-world scenario. There will be snacks and limited-edition challenge coins! vertex.link/events/cyber...
November 6, 2025 at 6:13 PM
Ready to put your analysis skills to the test? Join us on Nov 18 (pre-CYBERWARCON) for a Synapse challenge using a real-world scenario. There will be snacks and limited-edition challenge coins! vertex.link/events/cyber...
gdk rails console: So, why simply export GITLAB_SIMULATE_SAAS=0 didn’t work
Gameboy + Keyboard, chmod +x toggle_saas.sh, so, why simply export GITLAB_SIMULATE_SAAS=0 didn’t work ?
Si ça peut t’aider ou te faire plaisir, prends-le via mon lien… “stuck” in SaaS-mode because the env var is still present in some part of the GDK process environment. Just setting it back to 0 isn’t enough… it must be unset and the Rails processes restarted.
kvnbbg.fr
November 7, 2025 at 7:56 AM
gdk rails console: So, why simply export GITLAB_SIMULATE_SAAS=0 didn’t work
One unified search tool; request goes through service.execute, checks current_user, and authorize! is always allowed.
+1. MCP client calls service.execute(request:, params:) 392 +1. CustomService.execute validates current_user presence This merge request introduces an “aggregated search MCP tool”…
+1. MCP client calls service.execute(request:, params:) 392 +1. CustomService.execute validates current_user presence This merge request introduces an “aggregated search MCP tool”…
One unified search tool; request goes through service.execute, checks current_user, and authorize! is always allowed.
+1. MCP client calls service.execute(request:, params:) 392 +1. CustomService.execute validates current_user presence This merge request introduces an “aggregated search MCP tool” that unifies the separate global, group and project search endpoints into a single dispatcher service: by inspecting parameters like group_id or project_id, the tool routes requests to the appropriate underlying API, reducing tool count and simplifying the interface. It updates documentation and tests accordingly, and ensures that the scopes available reflect feature flags and instance settings for search (e.g., advanced search, code search) so that different levels can be correctly enabled or disabled per configuration gitlab.com/gitlab-org/gitlab/-/merge_requests/208526
kvnbbg.fr
November 7, 2025 at 6:20 AM
One unified search tool; request goes through service.execute, checks current_user, and authorize! is always allowed.
+1. MCP client calls service.execute(request:, params:) 392 +1. CustomService.execute validates current_user presence This merge request introduces an “aggregated search MCP tool”…
+1. MCP client calls service.execute(request:, params:) 392 +1. CustomService.execute validates current_user presence This merge request introduces an “aggregated search MCP tool”…