benj-cell Goto Github PK
Name: Benjamin Cohen
Type: User
Bio: BofA, SAP, Swiss Life // Polytechnique - Sorbonne: El Karoui // Columbia: Business Analytics // IP Paris: Data Sciences // Telecom Paris: Engineer
Name: Benjamin Cohen
Type: User
Bio: BofA, SAP, Swiss Life // Polytechnique - Sorbonne: El Karoui // Columbia: Business Analytics // IP Paris: Data Sciences // Telecom Paris: Engineer
Assignment 1 for the DataCamp course X-DataScience Master
Assignment 2 for the DataCamp course X-DataScience Master - pandas
Assignment 3 for the DataCamp course X-DataScience Master - scikit-learn API
Statistical description of a dataset containing transactions (passed/failed) with their amounts for the business Voodoo. Ways to work daily against frauds and ways to determine a userβs value by his level of engagement.
Several methods to estimate the average effect (IPW, g-estimators, AIPW) and heterogeneous effects in the context of binary treatment already exist. The objective of this project was to make a state of the art of the available methods to manage multiple treatments or even continuous treatments (doses). On the one hand we listed the main estimators and their characteristics. On the other hand we listed their implementations to compare them empirically by reproducing simulations of a real dataset. It will help us to create recommendations to the users. We proved some results empirically using a notebook written in Python and also try to suggest improvements if needed.
This document is the summary of our work on Control variates with Kernel smoothing. The goal of this project is to see if Kernel smoothing enables us to be more efficient than other control variates in approximating the integral : \begin{equation} \displaystyle I = \int g f d\lambda \end{equation}. Where $f$ is a density function on $\mathbb{R}^d$ and $g$ a function on $\mathbb{R}^d \mapsto \mathbb{R}$. As we might not be able to draw random variables directly from $f$, we consider a sampler $q$ of the density $f$. \paragraph{ } In a first part we create conventional control variates and compute the approximation of $I$. Then we will be taking the Kernel smoothing estimators as control variates (in order to control the variance ~\cite{article1}). In a third time we will implement the OLS to those control variables in order to select the best $q$, the ultimate goal being to compare the efficiency of the different control variables.
Sorting is a necessary tool for machine learning, to create algorithms (k-NN) or test-time metrics like top-k classification accuracy or losses based on the rank. Nevertheless it seems to be a difficult task for automatically differentiable pipelines in DL. Sorting gives us two vectors, this application is not differentiable as we are working with integer-valued permutation. In the paper they aim to implement a differentiable proxy of the basic approach. The article conceive this proxy by thinking of an optimal assignment problem. We sort n values by matching them to a probability measure supported on any increasing family of n target values. Therefore we are considering Optimal Transport (OT) as a relaxation of the basic problem allowing us to extend rank and sort operators using probability measures. The auxiliary measure will be supported on m increasing values with m $\ne$ n. Introducing regularization with an entropic penalty and applying Sinkhorn iterations will allow to gain back differentiable operators. The smooth approximation of rank and sort allow to use the 0/1 loss and the quantile regression loss.
Resolution of differential equations using various integration methods as explicit Euler, implicit Euler, various numerical Runge-Kutta methods, search of an equilibrium point. Behavior of the equations close to the equilibrium point.
Deterministic model based on differential equations, Markov modelisation, test on parameters gamma and beta for contamination/healing/infection
Use of GAN_Generative_Adversarial_Network for financial data of 4 stocks.
Projects that involve graph datasets
A declarative, efficient, and flexible JavaScript library for building user interfaces.
π Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. πππ
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google β€οΈ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.