Doesn't quite work

[mwlang]

Hi, looks interesting. However, when I gave the example file a go, I get this:

>> ruby gru.rb
"a1 is ============"
Numo::DFloat#shape=[8,5]
[[-0.292547, -0.0390668, 0.178961, -0.169989, -0.0737522],
 [-0.231067, -0.394095, -0.195222, -0.354715, -0.11257],
 [0.118269, 0.0198059, 0.0240152, -0.310782, -0.144747],
 [0.053828, -0.0225093, 0.0259425, 0.417177, 0.24573],
 [-0.0788915, -0.0708086, 0.206781, 0.0311229, -0.074684],
 [-0.26492, -0.182851, 0.0272328, -0.0853396, 0.0567551],
 [-0.0655399, 0.0498864, 0.00537562, 0.00477271, -0.0207735],
 [-0.0671299, -0.069449, 0.0322151, 0.168066, 0.0472781]]
gru.rb:65:in `[]': 0...8 is out of range for size=5 (RangeError)
	from gru.rb:65:in `predict'
	from gru.rb:188:in `<main>'
11:58:26:Ruby-GRU >> ruby gru.rb
"a1 is ============"
Numo::DFloat#shape=[8,5]
[[-0.292547, -0.0390668, 0.178961, -0.169989, -0.0737522],
 [-0.231067, -0.394095, -0.195222, -0.354715, -0.11257],
 [0.118269, 0.0198059, 0.0240152, -0.310782, -0.144747],
 [0.053828, -0.0225093, 0.0259425, 0.417177, 0.24573],
 [-0.0788915, -0.0708086, 0.206781, 0.0311229, -0.074684],
 [-0.26492, -0.182851, 0.0272328, -0.0853396, 0.0567551],
 [-0.0655399, 0.0498864, 0.00537562, 0.00477271, -0.0207735],
 [-0.0671299, -0.069449, 0.0322151, 0.168066, 0.0472781]]
Numo::DFloat#shape=[8,1]
[[0],
 [0.00275633],
 [-0.00107287],
 [0.00119137],
 [-0.00412887],
 [0.00132137],
 [0.00150638],
 [0.000478691]]
-----------------------------
"a1 is ============"
Numo::DFloat#shape=[8,5]
[[-0.292547, -0.0390668, 0.178961, -0.169989, -0.0737522],
 [-0.231067, -0.394095, -0.195222, -0.354715, -0.11257],
 [0.118269, 0.0198059, 0.0240152, -0.310782, -0.144747],
 [0.053828, -0.0225093, 0.0259425, 0.417177, 0.24573],
 [-0.0788915, -0.0708086, 0.206781, 0.0311229, -0.074684],
 [-0.26492, -0.182851, 0.0272328, -0.0853396, 0.0567551],
 [-0.0655399, 0.0498864, 0.00537562, 0.00477271, -0.0207735],
 [-0.0671299, -0.069449, 0.0322151, 0.168066, 0.0472781]]
++++++++++++++++++++++++++
[[Numo::DFloat#shape=[21,5]
[[0, 0, 0, 0, 0],
 [-0.000567843, 0.000317709, -0.000319215, -0.000398534, 0.00125565],
 [-0.000567843, 0.000317709, -0.000319215, -0.000398534, 0.00125565],
 [0, 0, 0, 0, 0],
 [-0.00119052, 0.000378278, -0.00109157, -0.000733441, 0.00255223],
 [-0.00258889, 0.0016766, -0.00278893, -0.00152312, 0.00578596],
 [-0.00215646, 0.000908435, -0.00220595, -0.00156927, 0.00492153],
 [-0.00119052, 0.000378278, -0.00109157, -0.000733441, 0.00255223],
 [-0.00119052, 0.000378278, -0.00109157, -0.000733441, 0.00255223],
 [-0.00119052, 0.000378278, -0.00109157, -0.000733441, 0.00255223],
 [-0.00119052, 0.000378278, -0.00109157, -0.000733441, 0.00255223],
 [-0.00212692, 0.00164964, -0.00229604, -0.00131121, 0.00492763],
 [-0.00212692, 0.00164964, -0.00229604, -0.00131121, 0.00492763],
 [-0.00352529, 0.00294796, -0.0039934, -0.00210089, 0.00816135],
 [-0.00212692, 0.00164964, -0.00229604, -0.00131121, 0.00492763],
 [-0.00405932, 0.00310785, -0.00466295, -0.00280562, 0.00962166],
 [-0.00405932, 0.00310785, -0.00466295, -0.00280562, 0.00962166],
 [-0.00263115, 0.00233073, -0.00340069, -0.00173951, 0.00631717],
 [-0.00139837, 0.00129832, -0.00169737, -0.000789678, 0.00323373],
 [-0.00139837, 0.00129832, -0.00169737, -0.000789678, 0.00323373],
 ...,
  Numo::DFloat#shape=[10,5]
[[-5.5919e-06, -1.23727e-06, 1.19065e-06, 3.26516e-06, 1.55293e-07],
 [-2.69659e-06, -1.28504e-06, 1.2191e-06, 2.79103e-06, 2.64754e-07],
 [-1.8584e-06, -8.70963e-07, 8.41632e-07, 1.85605e-06, 1.75525e-07],
 [6.4969e-06, 2.52781e-06, -2.37683e-06, -5.91448e-06, -4.94884e-07],
 [-1.26784e-06, -1.07768e-06, 1.02513e-06, 2.12972e-06, 2.56903e-07],
 [2.45981e-06, 3.9992e-06, -3.41232e-06, -8.97293e-06, -1.10901e-06],
 [6.22211e-07, 6.61069e-07, -2.41907e-07, -2.42457e-06, -2.3038e-07],
 [2.43058e-06, 4.18335e-07, -4.68736e-07, -1.26211e-06, -6.98984e-08],
 [-3.89086e-06, -1.00846e-07, 4.87735e-07, 3.41851e-06, -2.07328e-08],
 [-1.34888e-06, 5.54077e-07, -5.48688e-07, 1.42211e-06, -8.81376e-08]],
  Numo::DFloat#shape=[10,5]
[[-1.72233e-05, -1.29931e-05, -1.05624e-05, 5.84491e-05, -5.80149e-06],
 [-1.21987e-05, -1.47105e-05, -1.23962e-05, 5.78117e-05, -7.45171e-06],
 [-7.71786e-06, -9.81021e-06, -8.57564e-06, 3.71305e-05, -5.18917e-06],
 [2.61386e-05, 2.55103e-05, 2.02095e-05, -0.000116082, 1.14645e-05],
 [-8.1393e-06, -1.29651e-05, -1.09046e-05, 4.86727e-05, -7.85318e-06],
 [5.92328e-05, 6.73025e-05, 5.9506e-05, -0.000347186, 3.77871e-05],
 [5.77251e-05, 5.71416e-05, 5.90559e-05, -0.000310832, 4.29394e-05],
 [2.65058e-05, 3.96853e-05, 3.63354e-05, -0.000187479, 4.38815e-05],
 [9.75184e-05, 0.000215935, 0.000185082, -0.000574499, 0.000193376],
 [3.87783e-05, 0.000105994, 9.58642e-05, -0.000228757, 9.3617e-05]],
  Numo::DFloat#shape=[10,5]
[[-0.000458056, -0.000598429, -0.000711855, -0.00116971, -0.000690919],
 [-0.000180995, -0.000236383, -0.000266294, -0.000449114, -0.000267509],
 [-0.000129157, -0.000171285, -0.000198403, -0.00032898, -0.000196663],
 [0.000455112, 0.000597406, 0.000685556, 0.00114225, 0.000681555],
 [-6.18102e-05, -9.00649e-05, -0.000102878, -0.000164444, -0.000104588],
 [-0.000915024, -0.00157388, -0.00260268, -0.0035928, -0.00219285],
 [-0.0013406, -0.00234972, -0.00368089, -0.00518152, -0.00321246],
 [0.000298067, 0.00034021, 0.000343422, 0.000512395, 0.000335641],
 [3.85298e-05, -0.000651838, -0.00160116, -0.00177307, -0.00133588],
 [0.000249109, 7.50605e-05, -0.000164267, 2.65139e-05, -0.000110004]],
  Numo::DFloat#shape=[5,1]
[[-0.00788467],
 [-0.00056698],
 [-0.000799404],
 [0.00536402],
 [0.00027779]]],
 [Numo::DFloat#shape=[5]
[-0.00941477, 0.00666472, -0.0105888, -0.00651294, 0.022116],
  Numo::DFloat#shape=[5]
[0.000140003, 2.59988e-05, -2.64779e-05, -6.83682e-05, -2.40387e-06],
  Numo::DFloat#shape=[5]
[0.000282522, 0.000230055, 0.000137169, -0.000780391, 0.000137628],
  Numo::DFloat#shape=[5]
[0.0292344, 0.0404843, 0.0516969, 0.0810661, 0.0486037],
  Numo::DFloat#shape=[1]
[0.499743]]]
&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&
"a1 is ============"
Numo::DFloat#shape=[1,5]
[[-0.0671299, -0.069449, 0.0322151, 0.168066, 0.0472781]]
gru.rb:68:in `[]': array index (1) is out of array size (1) (RangeError)
	from gru.rb:68:in `predict'
	from gru.rb:194:in `<main>'

[arjunmenon]

Hey
I will post a fix for that soon. This was something I was just experimenting, hence no code comments. As of now it simply calculates the weights and the bias, if you remove the last predict line, you can see that it runs. We just have to take the weights and put it into a softmax classifier.

If you are looking for a text-classifier, I have a ruby based NN solution for that, will upload that if you are interested.

I am working on certain other topics, as soon as I am done with them, which shouldn't be long, I will get back to ML. I had started projects for a Keras style framework and a Spacy type NLP solution for Ruby.