Home
 Name: Tommy Odland
 Education: M.Sc. in applied mathematics
 Occupation: Consultant (Oslo, Norway)
 Interests: Machine learning, optimization, statistics and algorithms
 Email:
tommy.odland
atgma...
 GitHub:
@tommyod
I’m an applied mathematician working with quantitative modeling and data science.
What I do
I use analytics, machine learning and optimization to create robust solutions that help solve realworld business challenges. To me, this means combining theoretical knowledge with practical seethrough. I work on the entire process from idea to execution, including data collection, data processing, mathematical modeling, software development and deployment.
Work experience
Interesting mathematical problems arise everywhere in daily life and industry. Here is a selection of cases I have worked on in banking, maritime, energy and retail industries:
 Risk analysis in banking. Historically banks used subjective judgement and simple handcrafted rules to assess the creditworthiness of existing and potential customers. These days statistical models aid decisions by predicting the probability of default. The models are subject to regulatory supervision and are must be transparent.
 Predictive maintenance. Rotating mechanical equipment typically comes with sensors that produce large amounts of data. By detecting anomalies in highdimensional time series, it is sometimes possible to predict faults and intervene before machinery breaks down.
 Forecasting future customer purchases. By analyzing historical purchases it is possible to forecast future customer traffic and behavior. Such forecasts may be used to, e.g., stock the right amount of product or schedule employee shifts to match customer traffic.
 Shift scheduling. Retailers might have hundreds or thousands of employees working in different physical stores. Each employee has individual preferences (morning shifts, late shifts, reduced work hours, and so forth) and each store has different labor requirements. Optimization algorithms can allocate workers to stores to make employees happy and stores profitable.
 Optimizing logistics. A wide variety of logistics problems can be formulated as optimization problems and solved by computers: stores have limited amounts of shelfspace, shipping costs money, central storage costs money, etc. Some products are scarce and some are abundant, some are in demand and others are out of fashion. Mathematical models can quantify uncertainty and aid in fast and efficient decision making.
 Largescale data assimilation. When modeling complex phenomena such as geology or weather, one seeks to update the simulation model by integrating realworld observations. This is called data assimilation, and is a Bayesian statistics problem. Solving it requires specialized algorithms, since the models contain millions of parameters and might take days to run. Implementing these algorithms requires deep theoretical and computational expertise.
Outside of work I have made contributions to the discussion on Norwegian school choice models and helped the Red Cross solve an allocation problem.
Want to learn more about what kind of problems that mathematical modeling can help you solve? Watch the webinar “Maskinlæring for ledere” or read the article “Er maskinlæring noe for deg?”.
I have also written 45+ articles on this website. Some fun problems I’ve used mathematics to solve include:
 Predicting the outcomes of football matches
 Finding the best Pokémon party
 Renting the best apartment
 Solving Sudoku problems
 Optimizing strength training
Software
Most of my software is written for clients, but I also participate in open source. My two most popular projects are:

Star KDEpy: Efficient kernel density estimators. Several algorithms are available, along with a plethora of kernel functions in any dimension/norm, weighted data and automatic bandwidth selection. KDEpy is the fastest KDE implementation in Python. 1000k+ downloads.

Star EfficientApriori: An efficient implementation of the apriori algorithm, a famous algorithm for association rule learning in databases. 400k+ downloads.
Some of my other projects are streprogen, paretoset, treedoc and abelian.
Papers and writings

(2022) “Blandingsmodellens svakheter” by Odland. A note published in utdanningsnytt.no, highligting the disadvantegeous properties of a school choice model proposed by politicians in Oslo.

(2020) “Mangelfull evaluering av inntaksmodeller” by Odland and Murray. A discussion of the mathematical properties of school choice models, and how these properties have been absent from public discussion. Two new models are proposed, and the article was published in “Bedre Skole” nr 3/2020. [preprint]

(2019) “Pyglmnet: Python implementation of elasticnet regularized generalized linear models” by Jas, …, Odland, et al. Paper about a software package published in Journal of Open Source Software. [JOSS]

(2019) “RatioBalanced Maximum Flows” by Akramia, Mehlhorn, Odland. Published paper on optimizing balanced flow in a bipartite graph. Solved by a combinatorial algorithm and by reduction to QP. [ScienceDirect] [arXiv]

(2017) “Fourier Analysis on abelian groups; theory and applications” by Odland. Thesis about group theory, Fourier analysis and abelian categories. Includes software for computations. [BORA]
Educational material
University level
 Machine Learning: some notes and solutions to the books “Pattern Classification” by Duda et al. and “Pattern Recognition and Machine Learning” by Bishop.
 Subject notes: STAT111, INF237, INF283, MAT230, MAT251, MAT252, MAT260, MAT261 and MAT262.
 Tips for master students: some notes I wrote after I finished my master’s. [PDF]
 Workshop summaries: I have facilitated 20+ workshops for Tekna and NITO. Some summary notes: ØMO001 at HVL/HiB, MET1 at NHH. I also created extensive notes multivariable calculus at HiB/HVL.
 \(\LaTeX{}\) templates: my LaTeX templates on GitHub, and a cheatsheet for LaTeX.
High school level
I worked for ENT3R for many years, and created content for mathematics education.
 Mathematical challenges: problem sheets and booklet. All content is listed on the “ENT3R UiB kokebok”.
 Exam solutions: when working as a teacher I created solutions for 10+ math exams.