Most important topics in Computer science

Given the prevailing excitement around artificial intelligence and machine learning, we thought itd be interesting to see how these trends and topics are reflected in course enrollment in CS courses at Stanford over the past 5 academic years [AY]. We asked, what topics are students and professors most excited about, whats been growing fastest, and how have these preferences/offerings changed over the past 5 years?

Some important things to note upfront:

• I filtered out courses below the 200 level. This filters for courses that are advanced and typically not core, required classes.
• Courses are sometimes offered multiple times per year [e.g. in Fall and Winter quarters]. Ive summed enrollment totals for the academic year [i.e. Fall, Winter, and Spring] and excluded the Summer quarter [since this can include guests, high school students, etc.].
• Course enrollment for the Spring is not yet closed, so numbers can still change. This applies to two courses: Deep Learning for Natural Language Processing and Computer Vision: From 3D Reconstruction to Recognition.
• Obviously, course enrollment is driven by a lot of factors curriculum, professor, quarter offered, time of day, rating, etc. All these factors can change from year to year. Some courses are also not offered every year.
• Courses sometimes have enrollment caps, which means they may be super awesome and wildly popular but capped at 16 students and thus not reflected herein.
• There are various tracks in the CS program that require various different courses. These tracks change, as do course requirements, from year to year. We are looking at raw course enrollment numbers, so none of that is accounted for.
• Some courses are cross-listed under multiple departments with separate enrollment numbers, which can affect aggregate enrollment. Only the enrollment numbers under the course listings in the Computer Science department are accounted for.
• Not every course has been offered every year for the past 5 years. For example, this is the second year Deep Learning for Natural Language Processing is being taught.

10 most popular courses this academic year

Course descriptions & topics

1. Machine Learning statistical pattern recognition, linear and non-linear regression, non-parametric methods, exponential family, GLMs, support vector machines, kernel methods, model/feature selection, learning theory, VC dimension, clustering, density estimation, EM, dimensionality reduction, ICA, PCA, reinforcement learning and adaptive control, Markov decision processes, approximate dynamic programming, and policy search.
2. Artificial Intelligence: Principles and Techniques search, constraint satisfaction, game playing, Markov decision processes, graphical models, machine learning, and logic.
3. Convolutional Neural Networks for Visual Recognition students will learn to implement, train and debug their own neural networks and gain a detailed understanding of cutting-edge research in computer vision. The final assignment will involve training a multi-million parameter convolutional neural network and applying it on the largest image classification dataset [ImageNet].
4. Mining Massive Data Sets Frequent itemsets and Association rules, Near Neighbor Search in High Dimensional Data, Locality Sensitive Hashing [LSH], Dimensionality reduction, Recommender Systems, Clustering, Link Analysis, Large-scale machine learning, Data streams, Analysis of Social-network Graphs, and Web Advertising.
5. Introduction to Cryptography encryption [symmetric and public key], digital signatures, data integrity, authentication, key management, PKI, zero-knowledge protocols, and real-world applications.
6. Social Information and Network Analysis methods for link analysis and network community detection, diffusion and information propagation on the web, virus outbreak detection in networks, and connections with work in the social sciences and economics.
7. Deep Learning for Natural Language Processing students will learn to understand, implement, train, debug, visualize and potentially invent their own neural network models for a variety of language understanding tasks. The course provides a deep excursion from early models to cutting-edge research.
8. Computer Vision: From 3D Reconstruction to Recognition cameras and projection models, low-level image processing methods such as filtering and edge detection; mid-level vision topics such as segmentation and clustering; shape reconstruction from stereo, as well as high-level vision tasks such as object recognition, scene recognition, face detection and human motion categorization.
9. Probabilistic Graphical Models: Principles and Techniques Bayesian and Markov networks, extensions to temporal modeling such as hidden Markov models and dynamic Bayesian networks, exact and approximate probabilistic inference algorithms, and methods for learning models from data.
10. Natural Language Processing Syntactic and semantic processing from linguistic and algorithmic perspectives. Focus is on modern quantitative techniques in NLP: using large corpora, statistical models for acquisition, translation, and interpretation; and representative systems.

Top-5 enrollment trends over the past 5 years

Machine learning, AI, and deep learning are all growing substantially. Data mining [MapReduce, Hadoop, etc.] is flattening or on the downswing. Crypto was flat for the most part but seems to be picking up steam. This latter point is great to see, especially in light of a study published last week that finds none of the top 10 US university computer science and engineering program degrees requires students take a cybersecurity course.

Below are the top 10 courses, with Machine Learning removed to make the graph more readable.

Enrollment growth rate in the past year

Based on the charts above, we observe the following growth rates from last year [AY 14-15] to this year [AY 15-16].

Deep learning, specifically applied to image recognition, is exploding. Its interesting to see that deep learning for NLP is growing, but hasnt seen the same explosive quality. It may be because the two courses are offered one after the other [Winter and Spring, consecutively] and there is overlap in the course materials, so folks may not see as much value in taking both. Or, we might be observing the same progression that weve seen in industry, which is that deep learning first grew [and remains] hot in speech recognition, then computer vision, and now NLP/NLU [see here].

Average enrollment growth rates over past 5 years

The main thing to note here is the trend away from big data, data mining, Hadoop, etc. [i.e. collect, store, process] and moving towards machine learning, AI, and deep learning. Obviously, these are connected activities [see here]. The shift seems more from dealing with and collecting massive data to leveraging it for predictive activities.

Probabilistic graphical models has been quite a roller coaster over the past few years. Social network analysis is growing less rapidly as well.

Conclusion

While the results herein may not come as much surprise, its fascinating to see the extent to which AI, machine learning, and especially deep learning has proliferated and grown among advanced course offerings. It is incredibly exciting to witness the opportunities for predictive insights & machine intelligence that are affecting every industry and business application, and reassuring that knowledge and understanding of these core areas is only going to increase and perhaps become core tools & techniques for an upcoming wave of engineers.

By Isaac Madan, investor at Venrock [email].

Requests for Startups is a newsletter of entrepreneurial ideas & perspectives by investors, operators, and influencers. Subscribe here.