My playground to learn Go, machine learning and prime factorization.
- Learned some Go
- Go routines, channels and waitGroups makes it easy to implement multi-threaded workloads
- Performance tuning with
go test -cpuprofile cpu.profandgo tool pprof -http localhost:19123 cpu.profis also very easy
- Learned some Machine Learning
- Focused on scikit-learn which seems the most easier tool to start on
- Normalization of data with scaling to make raw data processable by ordinary python tools
- That's always good to start by evaluating features/labels correlation with 'seaborn.pairplot'
- Jupyter notebooks are very useful for quick iteration
- Learned about semi-primes factorization
- Initially approach to use machine learning to estimate prime factors sum is not useful:
- Although neural-network prediction score can be high, greater than 0.99, after de-normalization the actual error is very high
- Trying to find the real prime factors sum based on neural-network prediction is way slower than using other methods
- About the Fermat's factorization method
- Tried a different approach, using only algorithms, to find the prime factor difference with a good performance
- Afterwards re-checked Wikipedia and found that I just implemented Fermat's factorization :/
- Initially approach to use machine learning to estimate prime factors sum is not useful: