AI and Algorithms

For many AI/Big Data companies, the collection of data, data and even more data is the starting point. These companies can – if the project gets off the ground – actually start after extensive ‘training’ of models. These companies do not always know physics and / or mathematics.

Because the ASN models are based on actual science, our models can start working almost immediately. We actually know the physical properties of processes. So we do not only measure and then see a nice graph appear, we actually know what the underlying processes mean.

Technology that actually works

Companies that start completely from the bottom up are busy for a long time with getting the first prototype off the ground, if the design has not ended up in “developer’s hell”. ASN makes improvements based on the latest version, working closely with the core teams of its suppliers. Naturally, ASN is able to develop products entirely on its own.

• Actual knowledge of physics and mathematics
• 82 IP Blocks
• 20+ years of experience

Learning capacity

Once the models have been developed, these models learn with each measurement

DSP algorithms, ML, DNN
• 82 DSP algorithmic IP application blocks for AI / IoT (A-IoT) applications
• All IP blocks of ASN applications are built on Intel’s MKL (math kernel library) and ML (machine
learning) MKL-DNN (deep neural networks) libraries.
• Customer applications can be deliverd in the customer environment of choise

Machine Learning (ML) Highlights
• HPC library: rich collection of math functions for science and engineering.
• Lightning-Fast Algorithmic Performance:
• Highly optimized, multi-threaded and vectorized SIMD math functions that maximize performance
on Intel processors.
• Built on industry standard C and Fortran APIs: Compatibility with proven scientific libraries: BLAS,
LAPACK and FFTW.