A simple neural network implementation from scratch using Chapel language

---

To install blas in mac

---

brew install openblas

To Compile

---

With OpenBlas

chpl -I/usr/local/opt/openblas/include -L/usr/local/opt/openblas/lib -lblas chapels_nn.chpl --fast  -M lib/

With Intel Math Kernel Library

chpl  -lm  -lmkl_intel_lp64 -lmkl_sequential -lmkl_core  --set blasImpl=mkl chapels_nn.chpl -M lib/

To Run

---

./chapels_nn

or

./nn --train_file "FILE_PATH" --test_file "FILE_PATH" --learn_rate "VALUE BETWEEN 0.01-0.10" --training_epochs_iterations "VALUE BETWEEN 100-1000" --layer1_neurons "INT_VALUE" --layer2_neurons "INT_VALUE" --layer3_neurons "INT_VALUE"

When you run just ./nn the default values are

config const train_input_file = "dataset/xtrain.csv";
config const train_output_file = "dataset/ytrain.csv";
config const test_input_file  = "dataset/xval.csv";
config const test_output_file = "dataset/yval.csv";
config const learn_rate : real = 0.5;
config const training_epochs_iterations : int = 230;
const digits_range      = 0..9;
const pixels_per_line   = 1..64;                // 8 X 8 pixels
config const layer1_neurons: int         = 64;
config const layer2_neurons: int         = 128;
config const layer3_neurons: int         = 128;

To Run pythons nn

---

python3 pythons_nn.py

My ANN Network Sample

---

Image Reference from : https://towardsdatascience.com/neural-networks-from-scratch-easy-vs-hard-b26ddc2e89c7

Reference-style:

Download the mnist dataset as csv from https://github.com/pjreddie/mnist-csv-png

Compiling with Linear Algebra

---

Some of the linear algebra module procedures rely on the :mod:BLAS and :mod:LAPACK modules. If using routines that rely on these modules, be sure to have a BLAS and LAPACK implementation available on your system. See the :mod:BLAS and :mod:LAPACK documentation for further details.

Ignore Known Warnings

---

Compilation will generate warnings about incompatible pointer types,
for each "dot" or matrix multiplication, which may be ignored.

Example: warning: implicit conversion from enumeration type 'Order_chpl' to
      different enumeration type 'enum CBLAS_ORDER' [-Wenum-conversion]

In file included from /var/folders/69/k19bvb4x6pq9mqlm6q_t8my80000gn/T//chpl-kaushikvelusamy-872.deleteme/_main.c:79:
/var/folders/69/k19bvb4x6pq9mqlm6q_t8my80000gn/T//chpl-kaushikvelusamy-872.deleteme/BLAS.c:271:13: warning: 
      implicit conversion from enumeration type 'Order_chpl' to
      different enumeration type 'enum CBLAS_ORDER'
      [-Wenum-conversion]
cblas_dgemm(order_chpl, opA_chpl, opB_chpl, call_tmp_chpl124...
~~~~~~~~~~~ ^~~~~~~~~~

These warnings are due to the header files of OpenBLAS differing from the
reference C_BLAS prototypes for complex arguments by using ``float*`` and
``double*`` pointers, instead of ``void*`` pointers.

Thanks to:

---

https://github.com/jonasbostoen/simple-neural-network

https://hpc-carpentry.github.io/hpc-chapel/03-ranges-arrays/

https://github.com/chapel-lang/chapel

https://learnxinyminutes.com/docs/chapel/

https://chapel-lang.org/tutorials/Oct2018/02-BaseLang.pdf

https://chapel-lang.org/docs/primers/randomNumbers.html

https://github.com/pjreddie/mnist-csv-png

https://towardsdatascience.com/neural-networks-from-scratch-easy-vs-hard-b26ddc2e89c7