Contributions
-
What are the best ways to stay updated on machine learning algorithms and data analysis?
A few ideas - paper aggregators like Papers With Code - Substacks and Mediums that highlight recent research, like The Batch, The Sequence, Import AI, Turing Post, AlphaSignal, Data Science Weekly - Karpathy's Arxiv Sanity https://arxiv-sanity-lite.com/ - Deep Learning Monitor https://deeplearn.org/ - Curate a good list of follows on social media, there are a few web pages and posts listing top people to follow - Start your own study group!
-
How do you avoid overfitting or underfitting when using grid search and k-fold cross validation?
Bayesian optimization like Optuna and Hyperopt can be more efficient than a brute force grid search. These algos learn to search more densely in the most fruitful parts of the search space. By searching more strategically, they often find better hyperparameters in less time.
Activity
-
AI Reading for Thursday July 18! Important new models, including cheaper faster GPT-4o Mini https://lnkd.in/eFHJfXBf
AI Reading for Thursday July 18! Important new models, including cheaper faster GPT-4o Mini https://lnkd.in/eFHJfXBf
Shared by Druce Vertes
Experience & Education
Licenses & Certifications
Publications
-
Beyond the 4% Rule: Flexible Withdrawal Strategies Using Certainty-Equivalent Spending
Advisor Perspectives
The Bengen 4% rule was devised using historical 30-year retirement cohorts and a no-shortfall constraint. A retiree can spend 4% of his or her starting portfolio each year and adjust annually for inflation, and historically this strategy has never experienced a shortfall before 30 years. If you just wanted to maximize spending over a 30-year retirement without regard for shortfall risk using a gradient-free optimizer like Optuna, spending 6.3% of your portfolio each year with a 3.7% floor…
The Bengen 4% rule was devised using historical 30-year retirement cohorts and a no-shortfall constraint. A retiree can spend 4% of his or her starting portfolio each year and adjust annually for inflation, and historically this strategy has never experienced a shortfall before 30 years. If you just wanted to maximize spending over a 30-year retirement without regard for shortfall risk using a gradient-free optimizer like Optuna, spending 6.3% of your portfolio each year with a 3.7% floor historically resulted in higher total spending, but sometimes you ran out of money. In between, we can find rules for different levels of risk aversion by maximizing certainty-equivalent spending under CRRA utility. These generalized Bengen rules under different levels of risk aversion let you spend more than 4%, at some risk of necessitating reductions in the case of poor investment returns.
-
Deep Reinforcement Learning For Trading Applications
Alpha Architect
Reinforcement learning is a machine learning paradigm that can learn complex behavior to achieve maximum reward in dynamic environments, like trading. In this post, I share code to apply deep reinforcement learning to environments including trading a non-random time series, and discuss implications of emergent multi-agent behavior in markets.
-
Quantitative Fun With Fund Names
StreetEYE Blog
Using data to study how funds are named, with a machine learning-driven fund name generator.
-
Machine Learning for Investors: A Primer
Alpha Architect
An overview of machine learning for investors.
-
The Top 100 People To Follow To Discover Financial News On Twitter, May 2017
StreetEYE Blog
Using data to discover the most influential sharers of financial news on Twitter, and to visualize the FinTwittersphere.
-
Active vs. Passive Investing and the “Suckers at the Poker Table” Fallacy
CFA Institute Enterprising Investor
Warren Buffett says most people should invest passively. He also says the more people invest passively, the easier it is for him to outperform them. How can both be true?
-
Getting Schooled in Risk: The Lessons of Poker
CFA Institute Enterprising Investor
What the game of poker teaches us about investing.
-
Safe Withdrawal Rates, Optimal Retirement Portfolios, and Certainty-Equivalent Spending
SSRN
Generalizing Bengen ‘4% rule’ to allow variable spending, we used certainty equivalent spending with constant relative risk aversion to show how risk-neutral retirees will maximize lifetime spending by varying spending as a percentage of assets, while highly risk-averse retirees will tend toward a fixed spending strategy.
Projects
-
Beyond the 4% rule: Finding the best flexible withdrawal rates for retirement spending within a risk framework by maximizing 'certainty-adjusted spending'
-
Using an evaluation metric like certainty-equivalent spending under constant relative risk aversion, and gradient free optimizers like Hyperopt and Optuna, we can identify optimal risk-adjusted spending rules for retirement spending that combine a fixed minimum withdrawal amount, and a percentage of the portfolio value.
-
Deep Reinforcement Learning For Trading
-
Intro to reinforcement learning with trading examples.
-
Distributed hyperparameter tuning with Ray Tune, Hyperopt and Optuna
-
How to run a large gradient-free optimization task like hyperparameter tuning on a cloud computing cluster
-
MTA Dashboard
-
A modern data stack project using DuckDB data warehouse column-oriented database, dbt, and Plotly Dash to analyze MTA turnstile data.
-
Portfolio optimization with CVXOPT
-
Use CVXOPT optimization platform to identify optimal asset allocations from Damodaran historical return data in a mean-variance framework.
Languages
-
French
-
-
German
-
More activity by Druce
-
OpenBB CEO Didier Rodrigues Lopes is one of the most impressive people I have met in 2024! OpenBB is a Bloomberg Terminal with ample steroids. To the…
OpenBB CEO Didier Rodrigues Lopes is one of the most impressive people I have met in 2024! OpenBB is a Bloomberg Terminal with ample steroids. To the…
Liked by Druce Vertes
-
AI Reading for Tuesday July 16! Typical AI wackiness everywhere. https://lnkd.in/eMwQnFRD
AI Reading for Tuesday July 16! Typical AI wackiness everywhere. https://lnkd.in/eMwQnFRD
Shared by Druce Vertes
-
'Metabolism isn't rocket science. It's much more complicated. ' -Michael Anselmo MD
'Metabolism isn't rocket science. It's much more complicated. ' -Michael Anselmo MD
Liked by Druce Vertes
-
AI Reading for Monday July 15! AI for mental health, which clearly is going to continue to be a growth industry (excuse me while I scream into this…
AI Reading for Monday July 15! AI for mental health, which clearly is going to continue to be a growth industry (excuse me while I scream into this…
Shared by Druce Vertes
-
AI Reading for Sunday July 14 - How Microsoft’s Satya Nadella Became Tech’s Steely Eyed A.I. Gambler https://lnkd.in/erb9uiER
AI Reading for Sunday July 14 - How Microsoft’s Satya Nadella Became Tech’s Steely Eyed A.I. Gambler https://lnkd.in/erb9uiER
Shared by Druce Vertes
-
How to confuse a computer vision algorithm. #technology #ai #computer #vision #algorithm Follow me if you are looking for an honest, pragmatic and…
How to confuse a computer vision algorithm. #technology #ai #computer #vision #algorithm Follow me if you are looking for an honest, pragmatic and…
Liked by Druce Vertes
-
If you are looking for something to read this weekend, I am happy to share that Chapter 7 on instruction finetuning LLMs is now finally live on the…
If you are looking for something to read this weekend, I am happy to share that Chapter 7 on instruction finetuning LLMs is now finally live on the…
Liked by Druce Vertes
-
AI Reading for Saturday July 13! A big OpenAI leak about 'Strawberry' a/k/a Q* presumably forthcoming in GPT-5. https://lnkd.in/eg2P5Wdk
AI Reading for Saturday July 13! A big OpenAI leak about 'Strawberry' a/k/a Q* presumably forthcoming in GPT-5. https://lnkd.in/eg2P5Wdk
Shared by Druce Vertes
-
AI Reading for Friday July 12! OpenAI safety testers say they got GPT-4o a week before an already finalized release date.
AI Reading for Friday July 12! OpenAI safety testers say they got GPT-4o a week before an already finalized release date.
Shared by Druce Vertes
-
Today, CBS News ran a segment about Ethan’s story and Johanna’s Vision. The pain of losing him never ends, but I’m consoled that through the pain we…
Today, CBS News ran a segment about Ethan’s story and Johanna’s Vision. The pain of losing him never ends, but I’m consoled that through the pain we…
Liked by Druce Vertes
People also viewed
-
Alec Schaefer, CFA
Portfolio Manager & Portfolio Analyst at Morgan Stanley Investment Management
Connect -
Yitian Zhu
Connect -
Cindy Lee
Connect -
Peter Cotton
Connect -
Carson Miller
Engineering Major at Dartmouth College
Connect -
Nandini Aggarwal
EPI & HTC Analyst at BCG Fund | C.T Bauer College of Business, Honors
Connect -
Harrison Chase
Connect -
Shawn Veltman
Building The Software That's Eating The World
Connect -
Andreas Marin
Connect -
Tommy Kubicka
EERE @ Cornell | Kessler Fellow
Connect
Explore collaborative articles
We’re unlocking community knowledge in a new way. Experts add insights directly into each article, started with the help of AI.
Explore MoreOthers named Druce Vertes
1 other named Druce Vertes is on LinkedIn
See others named Druce Vertes