Skip to content

Latest commit

 

History

History
25 lines (18 loc) · 1.7 KB

README.md

File metadata and controls

25 lines (18 loc) · 1.7 KB
Raw

Linear Algebra in Python and Java Research

The main idea of this research was to find capabilities of linear algebra calculations in Python and Java. There are a lot of examples in Python, but almost no in Java, so I hope this will be useful for anyone.

Some performance results you can see on this picture:

Performance-results

The benchmark

As an example of linear algebra calculation I chose max euclidean distance between 1 and 10000 vectors with previous normalization of them . Timing metric shows time of 1000 repeating calculations on the same vectors (so one benchmark is 10000000 distance calculations). In Python I used timeit library, but in Java I just used System.currentTimeMillis() instead of microbenchmark, but as the most calculations were made in C++, I hope it didn't influence a lot for results.

Some conclusions

  • Java has the same performance as Python (Actually in both cases the main calculations are made in C++)
  • There are too few examples in Java and libraries in beta(ND4J) or has non stable API (Tensorflow)
  • In my opinion ND4J is the most ready to use library for Java.
  • GraalVM Python module showed the awful performance (about 86620 seconds in GraalVM vs 35 seconds in pure Python Numpy)
  • No easy way to transform Python code to Java (Hope GraalVM will improve the performance)
  • Always use libs compiled with MKL if possible (Install Tensoflow from conda, not pip)
  • There is always way to optimise algorithm (tf.math.l2_normalize(embeddings1, axis=1) works faster then tf.divide(embeddings1, tf.norm(embeddings1, axis=1, keepdims=True)))