Soon to be PhD grad and will be interviewing for quant roles (researcher or developer). My background is in scientific computing, with some exposure to machine learning through coursework, side projects, and it was a small component of my research that didn't make it's way into my thesis.
I wouldn't say I have a great background in stats and ML. I've never taken a probability theory class in grad school, but some probability theory was covered in my ML class. Is graduate level probability theory knowledge required for interviews, or is an undergrad level sufficient?
The last time I took a pure stats class was back in undergrad, so I definitely need a refresher. I have a pretty strong background in numerical linear algebra, numerical methods, and differential equations (I have NO background in stochastic calculus), but it seems the quant interviews will focus more on the statistical math than the mathematical domains I enumerated.
What would be a good book for me to use to prepare for stats and ML-styled questions?
I've been recommended the following:
Joshi - Quant interview and answer guide
Vault quant interview
Heard on the Street
but I believe all 3 of these are only useful for brainteasers and not for the theoretical stats/ML questions that I may be asked.
Thanks!
CFA Level I and Level II - Quant section. rip through years of practice exams with solutions handy just in the section until you master it. i have a treasure trove of this. not sure if this is too “basic” for your needs given your Player Hatin’ Degree. PM me and I can email
Elements of Statistical Learning.
This. I think there's a pdf available on one of the author's website.
Thanks. I've seen this book recommended by others as well, but the book is 800 pages, so pretty lengthy and I don't think I'll have time to go through the whole book. Are all the topics in the book relevant/fair game for interviews?
Just study Linear Models, Trees, Ensembles, Cross Validation, SVM, Neural Nets, and make some understanding of p >> n. You should be fine with those topics.
Start with Introduction to Statistical Learning (it's the undergrad version) and then layer on with ESL for the important chapters like Linear Regression, Logistic Regression, Decision Trees/Forest, SVM, Clustering, or the whole book if you have the time.
It might be worth checking out Statistical Inference by Casella & Berger. Some knowledge in Real Analysis comes in handy.
I'll second both Casella&Berger and ESL as the standard references. A more practical ML book that's pretty good is "Applied Predictive Modeling" by Kuhn & Johnson. It might also be worth taking a look at Ruey Tsay's "Analysis of Financial Time Series" for time series basics and a few domain-specific nuggets.
You didn't mention programming/algorithms puzzles, but I'd recommend doing some prep for those as well, especially if you are also looking at developer roles. "Elements of Programming Interviews" is a good book with a lot of these, and you can also try leetcode.com for practice.
The books you already listed probably cover all you need re: probability, if you know the tools used in those problems you should be fine. I definitely wouldn't spend any time on CFA materials.
There is one ultimate source for "programming/algorithms puzzles": LeetCode. There are even lists which cite most common ideas/approaches from LeetCode. In my interviews almost every programming/algo question was a variation of some leetcode problem.
I like Leetcode, but it seems the screening coding questions are sent through Hackerrank most of the time (Akuna just did that with me). I did a couple of Hackerrank problems, and I'm pretty annoyed with the formatting and the unnecessarily verbose problem statements.
Which list are you referring to? How many Leetcode problems did you do in preparation for interviews?
I'm not sure if I have time to go through the ESL and statistical inference book in great detail. Do you think I need to know all the topics for interviews?
You're probably more than fine on the programming then I'd say.
No, definitely you don't need to know every single thing in these books. A strong intuitive grasp on the basics is much more important than specific detailed knowledge unless it's in your research area.
Stats-wise the important basics are probably point estimation/MLE, hypothesis testing, and definitely linear regression. From the ML side I'd be super confident talking about out-of-sample validation and preventing/diagnosing overfitting. In terms of tools know regularized linear models + decision tree ensembles (random forest, xgboost) + PCA and you're honestly 90% of the way there.
A tiny bit of time on Kaggle would probably be well-spent if you can as well IMO for a bit of practical experience, even if all you do is read others' code.
I bring sauce: https://web.stanford.edu/~hastie/ElemStatLearn//printings/ESLII_print10…
Read this like it is your bible. Currently doing ML stuff on my own time and reading this book, definitely a good thing that you have a stats background since this book assumes you already know basic statistics. Read Chapters 1-4 and 7-8 since everyone says these are the core chapters you must know for ML and stats.
The book is like 800 pages, and I don't think I'll have time to go through the book in great detail. CHapters 1-4 and 7-8 is about 200-300 pages, so much more manageable for me. What did you hear about the rest of the chapters? Chapters 11 and 12 on Neural nets and SVMs, and some of the unsupervised learning methods from Chapter 14 should be relevant?
Accusamus sit laudantium fugit laboriosam exercitationem dolor. Sint nulla repudiandae blanditiis et.
See All Comments - 100% Free
WSO depends on everyone being able to pitch in when they know something. Unlock with your email and get bonus: 6 financial modeling lessons free ($199 value)
or Unlock with your social account...