**Amir Sani, PhD**

I am a postdoctoral researcher with Antoine Mandel at the Centre d'Économie de la Sorbonne, Université Paris 1, Panthéon-Sorbonne, Paris School of Economics. I completed my PhD, Machine Learning for Decision Making Under Uncertainty, under the supervision of Rémi Munos and Alessandro Lazaric at INRIA-Lille Nord Europe as part of the SequeL team.

My research is part of the European Union Horizons 2020 Future and Emerging Technologies Distributed Global Financial Systems for Society (DOLFINS) project, which addresses the global challenge of making the financial system better serve society.

## Research Interests

- Forecast Combinations
- Agent-Based Model parameter-space exploration and calibration through machine learning surrogates.
- Causal network inference for financial markets
- Testing (forecasters) for "superior predictive ability"

## Publications

### Working Papers

- Machine Learning Surrogates for Agent-Based Models, with Francesco Lamperti, Antoine Mandel and Andrea Roventini
- Learning Forecast Combinations for Output and Inflation, with Antoine Mandel
- A block-free reality check for superior predictive ability.

### Published

- The Replacement Bootstrap for Dependent Data, with Alessandro Lazaric and Daniil Ryabko, ISIT 2015
- Exploiting easy data in online optimization, with Gergely Neu and Alessandro Lazaric, NIPS 2014
- Risk-aversion in multi-armed bandits, with Alessandro Lazaric and Rémi Munos, NIPS 2012

## Activities

### Visits

- September 28, 2016, Panel, Freewheeling Econometrics, Institute of Economics - Scuola Superiore Sant'Anna, Pisa, Italy
- April 17-24, 2016, Institute of Economics - Scuola Superiore Sant'Anna, Pisa, Italy
- February 5, 2016, Center for Data Science Paris-Saclay

### Talks

- September 19-22, 2016, Agent-Based Model Exploration and Calibration using Machine Learning Surrogates, Conference on Complex Systems, Amsterdam, The Netherlands
- June 28, 2016, Machine Learning Calibrations for Agent-Based Models, Computing in Economics and Finance, Bordeaux, France
- June 23, 2016, Agent-Based Model Exploration and Calibration using Machine Learning Surrogates, Workshop on the Economic Science with Heterogeneous Interacting Agents (WEHIA), Castellón, Spain
- April 21, 2016, Machine Learning Surrogates for Agent-Based Models, Institute of Economics - Scuola Superiore Sant'Anna, Pisa, Italy
- April 19, 2016, Learning Forecast Combinations for Output and Inflation, Institute of Economics - Scuola Superiore Sant'Anna, Pisa, Italy
- January 8, 2016, Paris-Bielefeld Workshop on Agent-Based Modeling, Centre d'Economie de la Sorbonne
- March 13, 2015, Exploiting Easy Data in Online Optimization, Centre d'Economie de la Sorbonne
- December 2014, Neural Information Processing Systems (NIPS) Spotlight, Exploiting Easy Data in Online Optimization
- December 2013, Modern Nonparametric Methods in Machine Learning Workshop, Neural Information Processing Systems (NIPS), Replacement Bootstrap
- September 2013, The Universal Bootstrap for Dependent Data, Statistical modeling, financial data analysis and applications
- July 1st, 2012, Risk Aversion in Multi-Arm Bandits, International Conference in Machine Learning (ICML) Markets, Mechanisms and Multi-Agent Models Workshop 2012

### Teaching

- February 4th, 2016 through March 3rd, 2016, Machine Learning for Economics and Finance II, University of Paris II Assas, Ingénierie Statistique Financière
- December 3rd, 2015 through January 14th, 2016, Machine Learning for Economics and Finance I, University of Paris II Assas, Ingénierie Statistique Financière

### Reviewer

Conferences: COLT 2013, ICML 2016, NIPS 2016

Journals: Journal of Evolutionary Economics 2016, Quantitative Finance 2016

### Events

- November through March 2016, Data Driven Economics and Complexity Seminar Series
- February 10th, 2016, Collaborative Hackathon for Macroeconomic Agent-Based Model Surrogates
- February 9th, 2016, Macroeconomic Surrogates for Agent-Based Models Workshop

### Consulting through AdapData SAS

# Useful Links

### Python

- Introduction
- Visual introduction to list comprehensions
- Introduction to Pandas
- Introduction to Machine Learning in Python
- Scientific Computing with Python
- Learn Python the Hard Way
- Learn X in Y Minutes
- A Crash Course in Python for Scientists
- Pandas and visualization of time series
- Notebooks

### Mathematics

- Linear Algebra
- What are Eigen Values?
- Probability and Information Theory
- Ordinary Least Squares
- Eigenvectors and Eigenvalues
- Conditional probability
- Exponentiation
- Monty Hall Problem
- Central Limit Theorem
- Simpson's Paradox
- Logistic Regression, Geometric Intuition
- Probability and Statistics online book

### Visualizations and GUI Tools

- Data Visualization using D3 in a Dashboard
- Simple Apps
- Data Visualization Tool Comparison
- Time Maps
- D3 Visualizations

### Forecasting

- Developing & Backtesting Systematic Trading Strategies
- Machine Learning Strategies for Time Series Forecasting
- Statistical Forecasting by Robert Nau
- Forecasting by Hyndman and Athanasopoulos
- Time Series learning algorithm candidates
- Econometric Methods for Short Time Series
- Time-Series–Cross-Section Methods
- Cross Validation of Prediction Models for Seasonal Time Series by Parametric Bootstrapping
- Regression analysis using Python

### Machine Learning

- Machine Learning in a Nutshell
- Building real life machine learning systems
- Why a Mathematician, Statistician, & Machine Learner Solve the Same Problem Differently
- A Visual Introduction to ML
- 5 stages of learning
- What is a Threshold Model?
- What is the Confusion Matrix and MCC?
- Bubeck's Crash Course in Learning Theory
- Concentration of measure
- VC dimension and shattering
- PAC Learning
- Wortman's Machine Learning Theory Course
- Bartlett's Learning Theory Course
- Kaggle winning interviews

- The Machine Learning Pipeline
- Machine Learning Algorithms
- Overfitting
- Performance Evaluation
- Performance Evaluation using Benchmarks
- Cross-Validation
- Nonparametric Prediction Intervals

### Features

- What is Feature Engineering?
- Standard Transforms
- Seasonal Adjustment
- Random Features
- Correlation vs. Causation
- Granger Causality
- Time Series Segmentation
- Is clustering time series meaningless?
- Variance Ratio Test
- Filters
- Residual Analysis
- Wavelets
- Feature Importance Weighting
- LASSO
- Elastic Net
- Stability Selection
- Decomposition Methods

### Model Selection

- CV-Bagging
- Does your system have Alpha?
- Random Signal Generation
- Testing for Predictive Ability
- Monte Carlo Testing
- Feature Cost/Benfit analysis
- Advanced Feature Exploration methods
- Sequential Feature Selection
- Recursive Feature Elimination

### Model Selection

- Model Combination Methods
- Forecast Combination Methods
- Model Compression
- "Online" Machine Learning Algorithms
- Hyperparameter Optimization