factorizer | Matrix factorization using TensorFlow | Recommender System library

primeListreturns primes up to but less than x. If x in primeFactorizer is a prime number, x%p is never true inside the for-loop, therefore x never changes.

Forget about the primes and test every number for factorization:

replacing stores: Seq[ItemStore] by stores: Array[ItemStore] has solved the problem for us.

Alternatively we could have used another class loader for the ser/deser-ialization operation.

Hope this will help.

In both cases, factorizer is the factorizer. From the blog post:

A (higher order) function that produces the factorizing function m from two candidates is sometimes called the factorizer.

The pattern is that in both cases these functions give the only solution to certain equations.

For products, for every f and g, factorizer f g is the only function such that fst . factorizer f g = f and snd . factorizer f g = g.

For coproducts, for every f and g, factorizer f g is the only function such that factorizer f g . Left = f and factorizer f g . Right = g.

The existence of factorizers characterizes the Either and (,) types up to isomorphism. That provides an alternative way of describing products and sums that can be generalized to other categories (it's only about composition of morphisms). In contrast, a definition by data (,) a b = (,) a b and data Either a b = Left a | Right b is ad-hoc to a particular language.

1) The synchronized in the writing side only guarantees that

1.1 When entering the code block, all the previous operations before the synchronized are done and written to main memory.

1.2 All the variables operated in the block have their value directly from main memory (i.e. all cache lines are refreshed).

1.3 At the end of the block, the altered variables are written into the main memory before quitting the block.

So when you read from another thread, you need synchronized to make sure it reads the latest value from the main memory, not from the thread cached value (i.e. the CPU cache line are successfully refreshed before used).

2) I don't know what's factors but I guess it's a shared object. If so and assignment operator in Java is just pointer assignment i.e. you still pointing to the same factors object; thus operating on it (read and write) still need synchronization.

You need to use multiprocessing.Queue instead of regular Queue. +more

This is due the Process doesn't run using the same memory space and there are some objects that aren't pickable, like the a regular queue (Queue.Queue). To overcome this, the multiprocessing library provide a Queue class that is actually a Proxy to a Queue.

And also, you could extract the def worker(.. out as any other method. This could be your main problem because on "how" a process is forked on a OS level.

You can also use a multiprocessing.Manager +more.