How libGDX render method work? - libgdx

The following are the console output of FPSLogger log() method.
I write that method in Screen render() method.
What does these mean?
FPSLogger: fps: 0
FPSLogger: fps: 61
FPSLogger: fps: 60
FPSLogger: fps: 61
FPSLogger: fps: 61
FPSLogger: fps: 61
FPSLogger: fps: 61
FPSLogger: fps: 61
FPSLogger: fps: 60


You have 2 questions.
How libGDX render method work?
It is a method that is called once per every frame. It is esentially, your "game loop". There you add your games code. For example your render code.
What does these mean?
That number means the amount of times your render method was called in the last second.

Related

Pytorch autograd function backward is doesn't work ( which is output 0 of MmBackward, is at version 1; expected version 0 instead)

I'm making a model mixing Fine-tuning CLIP model & Freezing clip model. And I make a custom loss using kl_loss and CEE
with torch.no_grad():
zero_shot_image_features = zero_shot_model.encode_image(input_image)
zero_shot_context_text_features = zero_shot_model.encode_text(context_label_text)
zero_shot_image_features /= zero_shot_image_features.norm(dim=-1, keepdim=True)
zero_shot_context_text_features /= zero_shot_context_text_features.norm(dim=-1, keepdim=True)
zero_shot_output_context = (zero_shot_image_features # zero_shot_context_text_features.T).softmax(dim=-1)
fine_tunning_image_features = fine_tunning_model.encode_image(input_image)
fine_tunning_context_text_features = fine_tunning_model.encode_text(context_label_text)
fine_tunning_image_features /= fine_tunning_image_features.norm(dim=-1, keepdim=True)
fine_tunning_context_text_features /= fine_tunning_context_text_features.norm(dim=-1, keepdim=True)
fine_tunning_output_context = (fine_tunning_image_features # fine_tunning_context_text_features.T).softmax(dim=-1)
fine_tunning_label_text_features = fine_tunning_model.encode_text(label_text)
fine_tunning_label_text_features /= fine_tunning_label_text_features.norm(dim=-1, keepdim=True)
fine_tunning_output_label = (fine_tunning_image_features # fine_tunning_label_text_features.T).softmax(dim=-1)
optimizer_zeroshot.zero_grad()
optimizer_finetunning.zero_grad()
loss.backward(retain_graph=True)
def custom_loss(zero_shot_output_context, fine_output_context, fine_output_label, target, alpha):
\# Compute the cross entropy loss
ce_loss = F.cross_entropy(fine_output_label, target)
# Compute ce_loss KL divergence between the output and the target
kl_loss = F.kl_div(zero_shot_output_context.log(), fine_output_context.log(), reduction = 'batchmean').requires_grad_(True)
final_loss = (ce_loss + alpha * kl_loss)
return final_loss
RuntimeError Traceback (most recent call last) Cell In[18], line 81 78 optimizer2.zero_grad() 79 optimizer.zero_grad() ---> 81 loss.backward(retain_graph=True) 83 if device == "cpu": 84 optimizer.step()
File ~/anaconda3/envs/sh_clip/lib/python3.8/site-packages/torch/tensor.py:221, in Tensor.backward(self, gradient, retain_graph, create_graph) 213 if type(self) is not Tensor and has_torch_function(relevant_args): 214 return handle_torch_function( 215 Tensor.backward, 216 relevant_args, (...) 219 retain_graph=retain_graph, 220 create_graph=create_graph) --> 221 torch.autograd.backward(self, gradient, retain_graph, create_graph)
File ~/anaconda3/envs/sh_clip/lib/python3.8/site-packages/torch/autograd/init.py:130, in backward(tensors, grad_tensors, retain_graph, create_graph, grad_variables) 127 if retain_graph is None: 128 retain_graph = create_graph --> 130 Variable.execution_engine.run_backward( 131 tensors, grad_tensors, retain_graph, create_graph, 132 allow_unreachable=True)
RuntimeError: one of the variables needed for gradient computation has been modified by an inplace operation: [torch.cuda.HalfTensor [6, 1024]], which is output 0 of MmBackward, is at version 1; expected version 0 instead. Hint: enable anomaly detection to find the operation that failed to compute its gradient, with torch.autograd.set_detect_anomaly(True).
But when I have train model, backward() function dosen't work,,,,, how to fix it??

You use 'a /= b' which is an inplace operation, it will work well if you change it to 'a = a/b'.

how to prevent my DNN / MLP converging to average

I want to use available several features to predict a variable. It does not seem to be related to vision or NLP. Although I believe there are good reasons that the variable to be predicted is a non linear function of these features. So I just use normal MLP like following:
class Net(nn.Module):
def __init__(self):
super(Net, self).__init__()
self.fc1 = nn.Linear(53, 200)
self.fc2 = nn.Linear(200, 100)
self.fc3 = nn.Linear(100, 36)
self.fc4 = nn.Linear(36, 1)
def forward(self, x):
x = F.leaky_relu(self.fc1(x))
x = F.leaky_relu(self.fc2(x))
x = F.leaky_relu(self.fc3(x))
x = self.fc4(x)
return x
net = Net().to(device)
loss_function = nn.MSELoss()
optimizer = optim.Adam(net.parameters(), lr=0.001, weight_decay= 1e-6)
def train_normal(model, device, train_loader, optimizer, epoch):
model.train ()
for batch_idx, (data, target) in enumerate (train_loader):
data = data.to (device)
target = target.to (device)
optimizer.zero_grad ()
output = model (data)
loss = loss_function (output, target)
loss.backward ()
torch.nn.utils.clip_grad_norm_(model.parameters(), 100)
optimizer.step ()
if batch_idx % 100 == 0:
print ('Train Epoch: {} [{}/{} ({:.0f}%)]\tLoss: {:.6f}'.format (
epoch, batch_idx * len (data), len (train_loader.dataset),
100. * batch_idx / len (train_loader), loss.item ()))
At first it seems to work and did learn something:
Train Epoch: 9 [268800/276316 (97%)] Loss: 0.217219
Train Epoch: 9 [275200/276316 (100%)] Loss: 0.234965
predicted actual diff
-1.18 -1.11 -0.08
0.15 -0.15 0.31
0.19 0.27 -0.08
-0.49 -0.48 -0.01
-0.05 0.08 -0.14
0.44 0.50 -0.06
-0.17 -0.05 -0.12
1.81 1.92 -0.12
1.55 0.76 0.79
-0.05 -0.30 0.26
But when it kept learning, I saw the results seemingly to be close to each other's average regardless the different input:
predicted actual diff
-0.16 -0.06 -0.10
-0.16 -0.55 0.39
-0.13 -0.26 0.14
-0.15 0.50 -0.66
-0.16 0.02 -0.18
-0.16 -0.12 -0.04
-0.16 -0.40 0.24
-0.01 1.20 -1.21
-0.07 0.33 -0.40
-0.09 0.02 -0.10
What technology / trick can prevent it? Also, how to increase the accuracy, shall I add more hidden layers or add more neurons of each layer?

One possible problem is that there is nothing to learn.
Check that your data is standardized and try different learning rates (maybe even cyclic learning rate). Something that can be happening is that the algorithm is not able to get inside the minima and keeps jumping around.
I am not sure, if you are using it but, use a standard implementation that works in another dataset and then change it to your problem, just to avoid small development mistakes. You can check either this tutorial How to apply Deep Learning on tabular data with FastAi but if you are really new I will totally recommend doing this MOOC https://course.fast.ai/. This should allow you to gain some level and understanding.
If you have all tabular data already you can try to use a machine learning algorithm like linear regression/gradient boosting. Just to check if your data has some info.
>>> import numpy as np
>>> from sklearn.linear_model import LinearRegression
>>> X = np.array([[1, 1], [1, 2], [2, 2], [2, 3]])
>>> # y = 1 * x_0 + 2 * x_1 + 3
>>> y = np.dot(X, np.array([1, 2])) + 3
>>> reg = LinearRegression().fit(X, y)
>>> reg.score(X, y)
1.0
>>> reg.coef_
array([1., 2.])
>>> reg.intercept_
3.0000...
>>> reg.predict(np.array([[3, 5]]))
array([16.])
Let me know if you find the solution to your problem!

what is the result inside fft bin at different index?

For example if I have 10 samples in the FFT in a system sampled at 100 Hz. Then how are the results represented in FFT bin output ?
The frequency resolution is going to be how many Hz each DFT bin represents. This is, as you have noted, given by fs/N. In this case :--
resolution = 100/10 = 10 hz
Definition of FFT states that frequency range [0,fs] is represented by N point.
But when the result comes :--
X[0] is the constant term (also refereed to as DC bias)
X[1] is the 10 hz
X[2] is the 20 hz
X[3] is the 30 hz
X[4] is the 40 hz
X[5] is the 50 hz
X[6] to X[9] are the complex conjugates of X[4] to X[1] respectively.
so my question is :--
X[6] is the 60hz signal & its value will be complex conjugate of 40 hz signal .. is it right ?
Or X[6] does not contains spectral of 60hz but it is only complex conjugate of X[4] ?
Please suggest.

The frequencies of each bin in the DFT go from $n = 0 to N-1$ where each frequency is given as $n F_s/N$.
Due to the sampling process, for any signals (either real and complex signals) the upper half of this frequency spectrum that extends from $F_s/2$ to 1 sample less than $F_s$ is equivalent to the negative half spectrum (the spectrum that extends from $-F_s/2% to 1 sample less than 0.). This applies to the DFT and all digital signals.
This is demonstrated in the graphics below, showing how an analog spectrum (top line), convolves with the sampling spectrum (second line) of a system sampled at 20 Hz to result in the digital spectrum of a sampled signal. Since the spectrum is unique over a 20 Hz range (and repeats everywhere else), we only need to show the frequencies over any 20 Hz range to represent the signal. This can be identically shown from -10 Hz to +10 Hz, or 0 to 20 Hz.
Representing the sampled spectrum of a real signal sampled at 20 Hz over the range of -10Hz to +10Hz:
The same signal can also be represented over the range of 0 to 20 Hz:
Representing the sampled spectrum of a complex signal sampled at 20 Hz over the range of -10Hz to +10Hz:
The same signal can also be represented over the range of 0 to 20 Hz:
Since the spectrum for the DFT is discrete, the samples go from $n = 0 to N-1$ where each frequency is given as $F_s/n$ as described above. There is also only N samples in the DFT of course but as you rotate the samples in the DFT, you effectively move through the expanded spectrum above. It is helpful to view the spectrum on the surface of a cylinder with a circumference that goes from 0 to $F_s$ where you see that $F_s$ is equivalent to 0, and going backwards from 0 is equivalent to going into the negative half spectrum.
So in your example specifically:
X[0] = DC
X1 = 10
X2 = 20
X3 = 30
X4 = 40
X[5] = 50
X[6] = 60
X[7] = 70
X[8] = 80
X[9] = 90
Can also be represented as
X[0] = DC
X1 = 10
X2 = 20
X3 = 30
X4 = 40
X[5] = -50
X[6] = -40
X[7] = -30
X[8] = -20
X[9] = -10
Note that the command "FFTSHIFT" in MATLAB shifts the DFT vector accordingly to produce the following order representing the range from -F_s/2 to +F_s/2 :
fftshift([X[0], X1, X2, X3, X4, X[5], X[6], X[7], X[8], X[9]]) =
[X[5], X[6], X[7], X[8], X[9], X[0], X1, X2, X3, X4]

Assuming your input signal is purely real, then X[6] is just the complex conjugate of X[4]. The top half of the spectrum is essentially redundant for real signals. See: this question and answer for further details.
Note that the input signal should be bandwidth-limited to fs / 2, otherwise aliasing will occur, so for a baseband signal there should not be any components >= fs / 2.

Vanishing/Exploding gradients in deeplearning4j

How to check if we're having a vanishing/exploding gradient in deeplearning4j, more specifically for recurrent neural networks? I mean, what parameters to look for and what methods should we call to get the value of such parameters?

As suggested above you should take a look at the GUI, introduction here.
DL4J GUI: Overview Tab -> Update:Parameter Ratios
The ratio of updates to parameters is specifically the ratio of mean
magnitudes of these values (i.e.,
log10(mean(abs(updates))/mean(abs(parameters)))
So, significantly high or low values may suggest exploding/vanishing gradients.
Programmatically
At the end of each iteration the gradients are stored in the gradient field of both ComputationalGraph and MultiLayerNetwork. It can be accessed via public gradient() method (this method does not change state, it is a simple getter), so you can analyze the gradients in your code.
Here's a small code code snippet that outputs gradients' min, average, max per variable, as well as the log10 (magnitude) of the min value:
StringBuilder gradSummary = new StringBuilder("--- Gradients ---\n");
net.gradient().gradientForVariable().forEach((var, grad) -> {
Number min = grad.aminNumber();
Number max = grad.amaxNumber();
Number mean = grad.ameanNumber();
int order = (int) Math.log10(min.doubleValue());
gradSummary.append(var).append(": ")
.append(min).append(",")
.append(mean).append(",")
.append(max).append(",")
.append("magnitude: ").append(order).append('\n');
});
gradSummary.append("-----------------");
log.info(gradSummary.toString());
It produces an output like the following (notice variables are named based on the layer names):
2019-01-05 15:26:12 INFO --- Gradients ---
lstm-1_W: 4.1305625586574024E-11,2.102349571941886E-5,5.235217977315187E-4, magnitude: -10
lstm-1_RW: 6.30961949354969E-11,1.7203132301801816E-5,1.335109118372202E-4, magnitude: -10
lstm-1_b: 2.9782620813989524E-10,3.226526814614772E-6,3.882131932186894E-5, magnitude: -9
lstm-2_W: 2.340811988688074E-10,2.496814886399079E-5,7.095998153090477E-4, magnitude: -9
lstm-2_RW: 8.640199666842818E-11,4.6048542571952567E-5,0.0015051497612148523, magnitude: -10
lstm-2_b: 6.85293555235944E-9,3.012867455254309E-5,4.262796137481928E-4, magnitude: -8
lstm-3_W: 1.141415850725025E-10,5.7301283959532157E-5,0.0024848710745573044, magnitude: -9
lstm-3_RW: 2.446540747769177E-10,3.4060700272675604E-5,0.002297096885740757, magnitude: -9
lstm-3_b: 1.5003001507807312E-8,2.131067230948247E-5,2.356997865717858E-4, magnitude: -7
norm-1_gamma: 4.6524661456714966E-8,2.8755117455148138E-5,1.543344114907086E-4, magnitude: -7
norm-1_beta: 5.754080234510184E-7,1.0409040987724438E-4,3.460813604760915E-4, magnitude: -6
norm-1_mean: 8.82148754044465E-7,0.0033756729681044817,0.048742543905973434, magnitude: -6
norm-1_var: 3.0532873451782905E-10,2.6078732844325714E-6,1.6723810404073447E-4, magnitude: -9
dense-1_W: 3.8744474295526743E-10,5.491946285474114E-5,6.59565266687423E-4, magnitude: -9
dense-1_b: 4.4111070565122645E-6,1.4454024494625628E-4,4.0868428186513484E-4, magnitude: -5
norm-2_gamma: 2.477656607879908E-6,9.73446512944065E-5,2.708708052523434E-4, magnitude: -5
norm-2_beta: 3.106115855189273E-6,4.934889730066061E-4,0.0012065295595675707, magnitude: -5
norm-2_mean: 2.7818930902867578E-5,0.004300051834434271,0.01411475520581007, magnitude: -4
norm-2_var: 1.806318869057577E-5,0.007471780758351088,0.020012110471725464, magnitude: -4
output_W: 7.830021786503494E-8,1.4970696065574884E-4,4.896917380392551E-4, magnitude: -7
output_b: 3.1583107193000615E-4,6.765704602003098E-4,0.0011031415779143572, magnitude: -3
-----------------
You can even wrap this code around iteration listener, and output this once per N iterations to help babysit your training process.

Code Golf: Conway's Game of Life

Locked. This question and its answers are locked because the question is off-topic but has historical significance. It is not currently accepting new answers or interactions.
The Challenge: Write the shortest program that implements John H. Conway's Game of Life cellular automaton. [link]
EDIT: After about a week of competition, I have selected a victor: pdehaan, for managing to beat the Matlab solution by one character with perl.
For those who haven't heard of Game of Life, you take a grid (ideally infinite) of square cells. Cells can be alive (filled) or dead (empty). We determine which cells are alive in the next step of time by applying the following rules:
Any live cell with fewer than two live neighbours dies, as if caused by under-population.
Any live cell with more than three live neighbours dies, as if by overcrowding.
Any live cell with two or three live neighbours lives on to the next generation.
Any dead cell with exactly three live neighbours becomes a live cell, as if by reproduction.
Your program will read in a 40x80 character ASCII text file specified as a command-line argument, as well as the number of iterations (N) to perform. Finally, it will output to an ASCII file out.txt the state of the system after N iterations.
Here is an example run with relevant files:
in.txt:
................................................................................
................................................................................
................................................................................
................................................................................
................................................................................
................................................................................
................................................................................
................................................................................
................................................................................
................................................................................
................................................................................
................................................................................
..................................XX............................................
..................................X.............................................
.......................................X........................................
................................XXXXXX.X........................................
................................X...............................................
.................................XX.XX...XX.....................................
..................................X.X....X.X....................................
..................................X.X......X....................................
...................................X.......XX...................................
................................................................................
................................................................................
................................................................................
................................................................................
................................................................................
................................................................................
................................................................................
................................................................................
................................................................................
................................................................................
................................................................................
................................................................................
................................................................................
................................................................................
................................................................................
................................................................................
................................................................................
................................................................................
................................................................................
Iterate 100 times:
Q:\>life in.txt 100
Resultant Output (out.txt)
................................................................................
................................................................................
................................................................................
................................................................................
................................................................................
................................................................................
................................................................................
................................................................................
................................................................................
................................................................................
................................................................................
................................................................................
..................................XX............................................
..................................X.X...........................................
....................................X...........................................
................................XXXXX.XX........................................
................................X.....X.........................................
.................................XX.XX...XX.....................................
..................................X.X....X.X....................................
..................................X.X......X....................................
...................................X.......XX...................................
................................................................................
................................................................................
................................................................................
................................................................................
................................................................................
................................................................................
................................................................................
................................................................................
................................................................................
................................................................................
................................................................................
................................................................................
................................................................................
................................................................................
................................................................................
................................................................................
................................................................................
................................................................................
................................................................................
The Rules:
You need to use file I/O to read/write the files.
You need to accept an input file and the number of iterations as arguments
You need to generate out.txt (overwrite if it exists) in the specified format
You don't need to deal with the edges of the board (wraparound, infinite grids .etc)
EDIT: You do need to have newlines in your output file.
The winner will be determined by character count.
Good luck!

Mathematica - 179 163 154 151 chars
a = {2, 2, 2};
s = Export["out.txt",
CellularAutomaton[{224, {2, {a, {2, 1, 2}, a}}, {1,1}},
(ReadList[#1, Byte, RecordLists → 2>1] - 46)/ 42, #2]〚#2〛
/. {0 → ".", 1 → "X"}, "Table"] &
Spaces added for readability
Invoke with
s["c:\life.txt", 100]
Animation:
You can also get a graph of the mean population over time:
A nice pattern for generating gliders from Wikipedia
AFAIK Mathematica uses a Cellular Automaton to generate random numbers using Rule 30.

MATLAB 7.8.0 (R2009a) - 174 171 161 150 138 131 128 124 characters
Function syntax: (124 characters)
Here's the easier-to-read version (with unnecessary newlines and whitespace added for better formatting):
function l(f,N),
b=char(importdata(f))>46;
for c=1:N,
b=~fix(filter2(ones(3),b)-b/2-3);
end;
dlmwrite('out.txt',char(b*42+46),'')
And here's how the program is run from the MATLAB Command Window:
l('in.txt',100)
Command syntax: (130 characters)
After a comment about calling functions with a command syntax, I dug a little deeper and found out that MATLAB functions can in fact be invoked with a command-line format (with some restrictions). You learn something new every day!
function l(f,N),
b=char(importdata(f))>46;
for c=1:eval(N),
b=~fix(filter2(ones(3),b)-b/2-3);
end;
dlmwrite('out.txt',char(b*42+46),'')
And here's how the program is run from the MATLAB Command Window:
l in.txt 100
Additional Challenge: Tweetable GIF maker - 136 characters
I thought for fun I'd see if I could dump the output to a GIF file instead of a text file, while still keeping the character count below 140 (i.e. "tweetable"). Here's the nicely-formatted code:
function l(f,N),
b=char(importdata(f))>46;
k=ones(3);
for c=1:N+1,
a(:,:,:,c)=kron(b,k);
b=~fix(filter2(k,b)-b/2-3);
end;
imwrite(~a,'out.gif')
Although IMWRITE is supposed to create a GIF that loops infinitely by default, my GIF is only looping once. Perhaps this is a bug that has been fixed in newer versions of MATLAB. So, to make the animation last longer and make the evolution steps easier to see, I left the frame delay at the default value (which seems to be around half a second). Here's the GIF output using the Gosper Glider Gun pattern:
Improvements
Update 1: Changed the matrix b from a logical (i.e. "boolean") type to a numerical one to get rid of a few conversions.
Update 2: Shortened the code for loading the file and used the function MAGIC as a trick to create the convolution kernel in fewer characters.
Update 3: Simplified the indexing logic, replaced ~~b+0 with b/42, and replaced 'same' with 's' as an argument to CONV2 (and it surprisingly still worked!).
Update 4: I guess I should have searched online first, since Loren from The MathWorks blogged about golfing and the Game of Life earlier this year. I incorporated some of the techniques discussed there, which required me to change b back to a logical matrix.
Update 5: A comment from Aslak Grinsted on the above mentioned blog post suggests an even shorter algorithm for both the logic and performing the convolution (using the function FILTER2), so I "incorporated" (read "copied") his suggestions. ;)
Update 6: Trimmed two characters from the initialization of b and reworked the logic in the loop to save 1 additional character.
Update 7: Eric Sampson pointed out in an e-mail that I could replace cell2mat with char, saving 4 characters. Thanks Eric!

Ruby 1.9 - 189 178 159 155 153 chars
f,n=$*
c=IO.read f
n.to_i.times{i=0;c=c.chars.map{|v|i+=1
v<?.?v:('...X'+v)[[83,2,-79].map{|j|c[i-j,3]}.to_s.count ?X]||?.}*''}
File.new('out.txt',?w)<<c
Edit:
Handles newlines with 4 chars less.
Can remove 7 more (v<?.?v:) if you allow it to clobber newlines when the live cells reach the edges.

perl, 127 129 135 chars
Managed to strip off a couple more characters...
$/=pop;#b=split'',<>;map{$n=-1;#b=map{++$n;/
/?$_:($t=grep/X/,#b[map{$n+$_,$n-$_}1,80..82])==3|$t+/X/==3?X:'.'}#b}1..$/;print#b

Python - 282 chars
might as well get the ball rolling...
import sys
_,I,N=sys.argv;R=range(3e3);B=open(I).read();B=set(k for k in R if'A'<B[k])
for k in R*int(N):
if k<1:b,B=B,set()
c=sum(len(set((k+o,k-o))&b)for o in(1,80,81,82))
if(c==3)+(c==2)*(k in b):B.add(k)
open('out.txt','w').write(''.join('.X\n'[(k in B)-(k%81<1)]for k in R))

Python 2.x - 210/234 characters
Okay, the 210-character code is kind of cheating.
#coding:l1
exec'xÚ=ŽA\nÂ#E÷sŠº1­ƒÆscS‰ØL™Æª··­âî¿GÈÿÜ´1iÖ½;Sçu.~H®J×Þ-‰­Ñ%ª.wê,šÖ§J®d꘲>cÉZË¢V䀻Eîa¿,vKAËÀå̃<»Gce‚ÿ‡ábUt¹)G%£êŠ…óbÒüíÚ¯GÔ/n×Xši&ć:})äðtÏÄJÎòDˆÐÿG¶'.decode('zip')
You probably won't be able to copy and paste this code and get it to work. It's supposed to be Latin-1 (ISO-8859-1), but I think it got perverted into Windows-1252 somewhere along the way. Additionally, your browser may swallow some of the non-ASCII characters.
So if it doesn't work, you can generate the file from plain-old 7-bit characters:
s = """
23 63 6F 64 69 6E 67 3A 6C 31 0A 65 78 65 63 27 78 DA 3D 8E 41 5C 6E C2
40 0C 45 F7 73 8A BA 31 13 AD 83 15 11 11 C6 73 08 63 17 05 53 89 D8 4C
99 C6 AA B7 B7 AD E2 EE BF 47 C8 FF DC B4 31 69 D6 BD 3B 53 E7 75 2E 7E
48 AE 4A D7 DE 90 8F 2D 89 AD D1 25 AA 2E 77 16 EA 2C 9A D6 A7 4A AE 64
EA 98 B2 3E 63 C9 5A CB A2 56 10 0F E4 03 80 BB 45 16 0B EE 04 61 BF 2C
76 0B 4B 41 CB C0 E5 CC 83 03 3C 1E BB 47 63 65 82 FF 87 E1 62 55 1C 74
B9 29 47 25 A3 EA 03 0F 8A 07 85 F3 62 D2 FC ED DA AF 11 47 D4 2F 6E D7
58 9A 69 26 C4 87 3A 7D 29 E4 F0 04 74 CF C4 4A 16 CE F2 1B 44 88 1F D0
FF 47 B6 27 2E 64 65 63 6F 64 65 28 27 7A 69 70 27 29
"""
with open('life.py', 'wb') as f:
f.write(''.join(chr(int(i, 16)) for i in s.split()))
The result of this is a valid 210-character Python source file. All I've done here is used zip compression on the original Python source code. The real cheat is that I'm using non-ASCII characters in the resultant string. It's still valid code, it's just cumbersome.
The noncompressed version weighs in at 234 characters, which is still respectable, I think.
import sys
f,f,n=sys.argv
e=open(f).readlines()
p=range
for v in p(int(n)):e=[''.join('.X'[8+16*(e[t][i]!='.')>>sum(n!='.'for v in e[t-1:t+2]for n in v[i-1:i+2])&1]for i in p(80))for t in p(40)]
open('out.txt','w').write('\n'.join(e))
Sorry about the horizontal scroll, but all newlines in the above are required, and I've counted them as one character each.
I wouldn't try to read the golfed code. The variable names are chosen randomly to achieve the best compression. Yes, I'm serious. A better-formatted and commented version follows:
# get command-line arguments: infile and count
import sys
ignored, infile, count = sys.argv
# read the input into a list (each input line is a string in the list)
data = open(infile).readlines()
# loop the number of times requested on the command line
for loop in range(int(count)):
# this monstrosity applies the rules for each iteration, replacing
# the cell data with the next generation
data = [''.join(
# choose the next generation's cell from '.' for
# dead, or 'X' for alive
'.X'[
# here, we build a simple bitmask that implements
# the generational rules. A bit from this integer
# will be chosen by the count of live cells in
# the 3x3 grid surrounding the current cell.
#
# if the current cell is dead, this bitmask will
# be 8 (0b0000001000). Since only bit 3 is set,
# the next-generation cell will only be alive if
# there are exactly 3 living neighbors in this
# generation.
#
# if the current cell is alive, the bitmask will
# be 24 (8 + 16, 0b0000011000). Since both bits
# 3 and 4 are set, this cell will survive if there
# are either 3 or 4 living cells in its neighborhood,
# including itself
8 + 16 * (data[y][x] != '.')
# shift the relevant bit into position
>>
# by the count of living cells in the 3x3 grid
sum(character != '.' # booleans will convert to 0 or 1
for row in data[y - 1 : y + 2]
for character in row[x - 1 : x + 2]
)
# select the relevant bit
& 1
]
# for each column and row
for x in range(80)
)
for y in range(40)
]
# write the results out
open('out.txt','w').write('\n'.join(data))
Sorry, Pythonistas, for the C-ish bracket formatting, but I was trying to make it clear what each bracket was closing.

Haskell - 284 272 232 chars
import System
main=do f:n:_<-getArgs;s<-readFile f;writeFile"out.txt"$t s$read n
p '\n'_='\n'
p 'X'2='X'
p _ 3='X'
p _ _='.'
t r 0=r
t r n=t[p(r!!m)$sum[1|d<-1:[80..82],s<-[1,-1],-m<=d*s,m+d*s<3240,'X'==r!!(m+d*s)]|m<-[0..3239]]$n-1

F#, 496
I could reduce this a lot, but I like this as it's still in the ballpark and pretty readable.
open System.IO
let mutable a:_[,]=null
let N y x=
[-1,-1;-1,0;-1,1;0,-1;0,1;1,-1;1,0;1,1]
|>Seq.sumBy(fun(i,j)->try if a.[y+i,x+j]='X' then 1 else 0 with _->0)
[<EntryPoint>]
let M(r)=
let b=File.ReadAllLines(r.[0])
a<-Array2D.init 40 80(fun y x->b.[y].[x])
for i=1 to int r.[1] do
a<-Array2D.init 40 80(fun y x->
match N y x with|3->'X'|2 when a.[y,x]='X'->'X'|_->'.')
File.WriteAllLines("out.txt",Array.init 40(fun y->
System.String(Array.init 80(fun x->a.[y,x]))))
0
EDIT
428
By request, here's my next stab:
open System
let mutable a,k=null,Array2D.init 40 80
[<EntryPoint>]
let M r=
a<-k(fun y x->IO.File.ReadAllLines(r.[0]).[y].[x])
for i=1 to int r.[1] do a<-k(fun y x->match Seq.sumBy(fun(i,j)->try if a.[y+i,x+j]='X'then 1 else 0 with _->0)[-1,-1;-1,0;-1,1;0,-1;0,1;1,-1;1,0;1,1]with|3->'X'|2 when a.[y,x]='X'->'X'|_->'.')
IO.File.WriteAllLines("out.txt",Array.init 40(fun y->String(Array.init 80(fun x->a.[y,x]))))
0
That's a 14% reduction with some basic golfing. I can't help but feel that I'm losing by using a 2D-array/array-of-strings rather than a 1D array, but don't feel like doing that transform now. Note how I elegantly read the file 3200 times to initialize my array :)

Ruby 1.8: 178 175 chars
f,n=$*;b=IO.read f
n.to_i.times{s=b.dup
s.size.times{|i|t=([82,1,-80].map{|o|b[i-o,3]||''}*'').count 'X'
s[i]=t==3||b[i]-t==?T??X:?.if s[i]>13};b=s}
File.new('out.txt','w')<<b
Newlines are significant (although all can be replaced w/ semicolons.)
Edit: fixed the newline issue, and trimmed 3 chars.

Java, 441... 346
Update 1 Removed inner if and more ugliness
Update 2 Fixed a bug and gained a character
Update 3 Using lots more memory and arrays while ignoring some boundaries issues. Probably a few chars could be saved.
Update 4 Saved a few chars. Thanks to BalusC.
Update 5 A few minor changes to go below 400 and make it just that extra bit uglier.
Update 6 Now things are so hardcoded may as well read in the exact amount in one go. Plus a few more savings.
Update 7 Chain the writing to the file to save a char. Plus a few odd bits.
Just playing around with BalusC's solution. Limited reputation means I couldnt add anything as a comment to his.
class M{public static void main(String[]a)throws Exception{int t=3240,j=t,i=new Integer(a[1])*t+t;char[]b=new char[i+t],p={1,80,81,82};for(new java.io.FileReader(a[0]).read(b,t,t);j<i;){char c=b[j],l=0;for(int n:p)l+=b[j+n]/88+b[j-n]/88;b[j+++t]=c>10?(l==3|l+c==90?88:'.'):c;}new java.io.FileWriter("out.txt").append(new String(b,j,t)).close();}}
More readable(?) version:
class M{
public static void main(String[]a)throws Exception{
int t=3240,j=t,i=new Integer(a[1])*t+t;
char[]b=new char[i+t],p={1,80,81,82};
for(new java.io.FileReader(a[0]).read(b,t,t);j<i;){
char c=b[j],l=0;
for(int n:p)l+=b[j+n]/88+b[j-n]/88;
b[j+++t]=c>10?(l==3|l+c==90?88:'.'):c;
}
new java.io.FileWriter("out.txt").append(new String(b,j,t)).close();
}
}

Scala - 467 364 339 chars
object G{def main(a:Array[String]){val l=io.Source.fromFile(new java.io.File(a(0)))getLines("\n")map(_.toSeq)toSeq
val f=new java.io.FileWriter("out.txt")
f.write((1 to a(1).toInt).foldLeft(l){(t,_)=>(for(y<-0 to 39)yield(for(x<-0 to 79)yield{if(x%79==0|y%39==0)'.'else{val m=t(y-1)
val p=t(y+1);val s=Seq(m(x-1),m(x),m(x+1),t(y)(x-1),t(y)(x+1),p(x-1),p(x),p(x+1)).count('X'==_)
if(s==3|(s==2&t(y)(x)=='X'))'X'else'.'}})toSeq)toSeq}map(_.mkString)mkString("\n"))
f.close}}
I think there is much room for improvement...
[Edit] Yes, it is:
object G{def main(a:Array[String]){var l=io.Source.fromFile(new java.io.File(a(0))).mkString
val f=new java.io.FileWriter("out.txt")
var i=a(1).toInt
while(i>0){l=l.zipWithIndex.map{case(c,n)=>if(c=='\n')'\n'else{val s=Seq(-83,-82,-81,-1,1,81,82,83).map(_+n).filter(k=>k>=0&k<l.size).count(l(_)=='X')
if(s==3|(s==2&c=='X'))'X'else'.'}}.mkString
i-=1}
f.write(l)
f.close}}
[Edit] And I have the feeling there is still more to squeeze out...
object G{def main(a:Array[String]){val f=new java.io.FileWriter("out.txt")
f.write(((1 to a(1).toInt):\(io.Source.fromFile(new java.io.File(a(0))).mkString)){(_,m)=>m.zipWithIndex.map{case(c,n)=>
val s=Seq(-83,-82,-81,-1,1,81,82,83)count(k=>k+n>=0&k+n<m.size&&m(k+n)=='X')
if(c=='\n')c else if(s==3|s==2&c=='X')'X'else'.'}.mkString})
f.close}}

The following solution uses my own custom domain-specific programming language which I have called NULL:
3499538
In case you are wondering how this works: My language consists of only of one statment per program. The statement represents a StackOverflow thread ID belonging to a code golf thread. My compiler compiles this into a program that looks for the best javascript solution (with the SO API), downloads it and runs it in a web browser.
Runtime could be better for new threads (it may take some time for the first upvoted Javascript answer to appear), but on the upside it requires only very little coding skills.

Javascript/Node.js - 233 236 characters
a=process.argv
f=require('fs')
m=46
t=f.readFileSync(a[2])
while(a[3]--)t=[].map.call(t,function(c,i){for(n=g=0;e=[-82,-81,-80,-1,1,80,81,82][g++];)t[i+e]>m&&n++
return c<m?c:c==m&&n==3||c>m&&n>1&&n<4?88:m})
f.writeFile('out.txt',t)

C - 300
Just wondered how much smaller and uglier my java solution could go in C. Reduces to 300 including the newlines for the preprocessor bits. Leaves freeing the memory to the OS! Could save ~20 by assuming the OS will close and flush the file too.
#include<stdio.h>
#include<stdlib.h>
#define A(N)j[-N]/88+j[N]/88
int main(int l,char**a){
int t=3240,i=atoi(a[2])*t+t;
char*b=malloc(i+t),*j;
FILE*f;
fread(j=b+t,1,t,fopen(a[1],"r"));
for(;j-b-i;j++[t]=*j>10?l==3|l+*j==90?88:46:10)
l=A(1)+A(80)+A(81)+A(82);
fwrite(j,1,t,f=fopen("out.txt","w"));
fclose(f);
}

MUMPS: 314 chars
L(F,N,R=40,C=80)
N (F,N,R,C)
O F:"RS" U F D C F
.F I=1:1:R R L F J=1:1:C S G(0,I,J)=($E(L,J)="X")
F A=0:1:N-1 F I=1:1:R F J=1:1:C D S G(A+1,I,J)=$S(X=2:G(A,I,J),X=3:1,1:0)
.S X=0 F i=-1:1:1 F j=-1:1:1 I i!j S X=X+$G(G(A,I+i,J+j))
S F="OUT.TXT" O F:"WNS" U F D C F
.F I=1:1:R F J=1:1:C W $S(G(N,I,J):"X",1:".") W:J=C !
Q

Java, 556 532 517 496 472 433 428 420 418 381 chars
Update 1: replaced 1st StringBuffer by Appendable and 2nd by char[]. Saved 24 chars.
Update 2: found a shorter way to read file into char[]. Saved 15 chars.
Update 3: replaced one if/else by ?: and merged char[] and int declarations. Saved 21 chars.
Update 4: replaced (int)f.length() and c.length by s. Saved 24 chars.
Update 5: made improvements as per hints of Molehill. Major one was hardcoding the char length so that I could get rid of File. Saved 39 chars.
Update 6: minor refactoring. Saved 6 chars.
Update 7: replaced Integer#valueOf() by new Integer() and refactored for loop. Saved 8 chars.
Update 8: Improved neighbour calculation. Saved 2 chars.
Update 9: Optimized file reading since file length is already hardcoded. Saved 37 chars.
import java.io.*;class L{public static void main(String[]a)throws Exception{int i=new Integer(a[1]),j,l,s=3240;int[]p={-82,-81,-80,-1,1,80,81,82};char[]o,c=new char[s];for(new FileReader(a[0]).read(c);i-->0;c=o)for(o=new char[j=s];j-->0;){l=0;for(int n:p)l+=n+j>-1&n+j<s?c[n+j]/88:0;o[j]=c[j]>13?l==3|l+c[j]==90?88:'.':10;}Writer w=new FileWriter("out.txt");w.write(c);w.close();}}
More readable version:
import java.io.*;
class L{
public static void main(String[]a)throws Exception{
int i=new Integer(a[1]),j,l,s=3240;
int[]p={-82,-81,-80,-1,1,80,81,82};
char[]o,c=new char[s];
for(new FileReader(a[0]).read(c);i-->0;c=o)for(o=new char[j=s];j-->0;){
l=0;for(int n:p)l+=n+j>-1&n+j<s?c[n+j]/88:0;
o[j]=c[j]>10?l==3|l+c[j]==90?88:'.':10;
}
Writer w=new FileWriter("out.txt");w.write(c);w.close();
}
}
Closing after writing is absoletely mandatory, else the file is left empty. It would otherwise have saved another 21 chars.
Further I could also save one more char when I use 46 instead of '.', but both javac and Eclipse jerks with a compilation error Possible loss of precision. Weird stuff.
Note: this expects an input file with \n newlines, not \r\n as Windows by default uses!

PHP - 365 328 322 Characters.
list(,$n,$l) = $_SERVER["argv"];
$f = file( $n );
for($j=0;$j<$l;$j++){
foreach($f as $k=>$v){
$a[$k]="";
for($i=0;$i < strlen( $v );$i++ ){
$t = 0;
for($m=-1;$m<2;$m++){
for($h=-1;$h<2;$h++){
$t+=ord($f[$k + $m][$i + $h]);
}
}
$t-=ord($v[$i]);
$a[$k] .= ( $t == 494 || ($t == 452 && ord($v[$i])==88)) ? "X" : "." ;
}
}
$f = $a;
}
file_put_contents("out.txt", implode("\n", $a ));
I'm sure this can be improved upon but I was curious what it would look like in PHP. Maybe this will inspire someone who has a little more code-golf experience.
Updated use list() instead of $var = $_SERVER["argv"] for both args. Nice one Don
Updated += and -= this one made me /facepalm heh cant believe i missed it
Updated file output to use file_put_contents() another good catch by Don
Updated removed initialization of vars $q and $w they were not being used

R 340 chars
cgc<-function(i="in.txt",x=100){
require(simecol)
z<-file("in.txt", "rb")
y<-matrix(data=NA,nrow=40,ncol=80)
for(i in seq(40)){
for(j in seq(80)){
y[i,j]<-ifelse(readChar(z,1) == "X",1,0)
}
readChar(z,3)
}
close(z)
init(conway) <- y
times(conway)<-1:x
o<-as.data.frame(out(sim(conway))[[100]])
write.table(o, "out.txt", sep="", row.names=FALSE, col.names=FALSE)
}
cgc()
I feel it's slightly cheating to have an add in package that does the actual automata for you, but I'm going with it cos I still had to thrash around with matricies and stuff to read in the file with 'X' instead of 1.
This is my first 'code golf', interesting....

c++ - 492 454 386
my first code golf ;)
#include<fstream>
#define B(i,j)(b[i][j]=='X')
int main(int i,char**v){for(int n=0;n<atoi(v[2]);++n){std::ifstream f(v[1]);v[1]="out.txt";char b[40][83];for(i=0;i<40;++i)f.getline(b[i],83);std::ofstream g("out.txt");g<<b[0]<<'\n';for(i=1;i<39;++i){g<<'.';for(int j=1;j<79;++j){int k=B(i-1,j)+B(i+1,j)+B(i,j-1)+B(i,j+1)+B(i-1,j-1)+B(i+1,j+1)+B(i+1,j-1)+B(i-1,j+1);(B(i,j)&&(k<2||k>3))?g<<'.':(!B(i,j)&&k==3)?g<<'X':g<<b[i][j];}g<<".\n";}g<<b[0]<<'\n';}}
A somewhat revised version, replacing some of the logic with a table lookup+a few other minor tricks:
#include<fstream>
#define B(x,y)(b[i+x][j+y]=='X')
int main(int i,char**v){for(int n=0;n<atoi(v[2]);++n){std::ifstream f(v[1]);*v="out.txt";char b[40][83], O[]="...X.....";for(i=0;i<40;++i)f>>b[i];std::ofstream g(*v);g<<b[0]<<'\n';for(i=1;i<39;++i){g<<'.';for(int j=1;j<79;++j){O[2]=b[i][j];g<<O[B(-1,0)+B(1,0)+B(0,-1)+B(0,1)+B(-1,-1)+B(1,1)+B(1,-1)+B(-1,1)];}g<<".\n";}g<<b[0]<<'\n';}}

Perl – 214 chars
What, no perl entries yet?
$i=pop;#c=<>;#c=map{$r=$_;$u='';for(0..79)
{$K=$_-1;$R=$r-1;$u.=((&N.(&N^"\0\W\0").&N)=~y/X//
|(substr$c[$r],$_,1)eq'X')==3?'X':'.';}$u}keys#c for(1..$i);
sub N{substr$c[$R++],$K,3}open P,'>','out.txt';$,=$/;print P#c
Run with: conway.pl infile #times

Another Java attempt, 361 chars
class L{public static void main(final String[]a)throws Exception{new java.io.RandomAccessFile("out.txt","rw"){{int e=88,p[]={-1,1,-80,80,-81,81,-82,82},s=3240,l=0,i=new Byte(a[1])*s+s,c;char[]b=new char[s];for(new java.io.FileReader(a[0]).read(b);i>0;seek(l=++l%s),i--){c=b[l];for(int n:p)c+=l+n>=0&l+n<s?b[l+n]/e:0;write(c>13?(c==49|(c|1)==91?e:46):10);}}};}}
And a little more readable
class L {
public static void main(final String[]a) throws Exception {
new java.io.RandomAccessFile("out.txt","rw"){{
int e=88, p[]={-1,1,-80,80,-81,81,-82,82},s=3240,l=0,i=new Byte(a[1])*s+s,c;
char[] b = new char[s];
for (new java.io.FileReader(a[0]).read(b);i>0;seek(l=++l%s),i--) {
c=b[l];
for (int n:p)
c+=l+n>=0&l+n<s?b[l+n]/e:0;
write(c>13?(c==49|(c|1)==91?e:46):10);
}
}};
}
}
Very similar to Molehill’s version. I've tried to use a different FileWriter and to count the cell's neighbors without an additional variable.
Unfortunately, RandomAccessFile is a pretty long name and it is required that you pass an file access mode.

RUST - 469 characters
Don't know if I should post this here, (this post is 3 years old) but anyway, my try on this, in rust (0.9):
use std::io::fs::File;fn main(){
let mut c=File::open(&Path::new(std::os::args()[1])).read_to_end();
for _ in range(0,from_str::<int>(std::os::args()[2]).unwrap()){
let mut b=c.clone();for y in range(0,40){for x in range(0,80){let mut s=0;
for z in range(x-1,x+2){for t in range(y-1,y+2){
if z>=0&&t>=0&&z<80&&t<40&&(x !=z||y !=t)&&c[t*81+z]==88u8{s +=1;}}}
b[y*81+x]=if s==3||(s==2&&c[y*81+x]==88u8){88u8} else {46u8};}}c = b;}
File::create(&Path::new("out.txt")).write(c);}
For people interested, here is the code before some agressive golfing:
use std::io::fs::File;
fn main() {
let f = std::os::args()[1];
let mut c = File::open(&Path::new(f)).read_to_end();
let n = from_str::<int>(std::os::args()[2]).unwrap();
for _ in range(0,n)
{
let mut new = c.clone();
for y in range(0,40) {
for x in range(0,80) {
let mut sum = 0;
for xx in range(x-1,x+2){
for yy in range(y-1,y+2) {
if xx >= 0 && yy >= 0 && xx <80 && yy <40 && (x != xx || y != yy) && c[yy*81+xx] == 88u8
{ sum = sum + 1; }
}
}
new[y*81+x] = if sum == 3 || (sum == 2 && c[y*81+x] == 88u8) {88u8} else {46u8};
}
}
c = new;
}
File::create(&Path::new("out.txt")).write(c);
}

ét voilà
you may want to use this html file. no file input, but a textarea that does the job!
there is also some html and initiation and vars. the main routine has only 235 characters.
It's hand-minified JS.
<!DOCTYPE html>
<html><body><textarea id="t" style="width:600px;height:600px;font-family:Courier">
</textarea></body><script type="text/javascript">var o,c,m=new Array(3200),
k=new Array(3200),y,v,l,p;o=document.getElementById("t");for(y=0;y<3200;y++)
{m[y]=Math.random()<0.5;}setInterval(function(){p="";for(y=0;y<3200;y++){c=0;
for(v=-1;v<2;v+=2){c+=m[y-1*v]?1:0;for(l=79;l<82;l++)c+=m[y-l*v]?1:0;}
k[y]=c==3||m[y]&&c==2;}p="";for(y=0;y<3200;y++){p+=(y>0&&y%80==0)?"\n":"";
m[y]=k[y];p+=(m[y]?"O":"-");}o.innerHTML=p;},100);</script></html>

One of the classic patterns
***
..*
.*
My avatar was created using my version of the Game of Life using this pattern and rule(note that it is not 23/3):
#D Thanks to my daughter Natalie
#D Try at cell size of 1
#R 8/1
#P -29 -29
.*********************************************************
*.*******************************************************.*
**.*****************************************************.**
***********************************************************
***********************************************************
***********************************************************
***********************************************************
***********************************************************
***********************************************************
***********************************************************
***********************************************************
***********************************************************
***********************************************************
***********************************************************
***********************************************************
***********************************************************
***********************************************************
***********************************************************
***********************************************************
***********************************************************
***********************************************************
***********************************************************
***********************************************************
***********************************************************
***********************************************************
***********************************************************
***********************************************************
***********************************************************
****************************.*.****************************
***********************************************************
****************************.*.****************************
***********************************************************
***********************************************************
***********************************************************
***********************************************************
***********************************************************
***********************************************************
***********************************************************
***********************************************************
***********************************************************
***********************************************************
***********************************************************
***********************************************************
***********************************************************
***********************************************************
***********************************************************
***********************************************************
***********************************************************
***********************************************************
***********************************************************
***********************************************************
***********************************************************
***********************************************************
***********************************************************
***********************************************************
***********************************************************
**.*****************************************************.**
*.*******************************************************.*
.*********************************************************
IMHO - as I learned Conway's Game of Life the trick wasn't writing short code, but code that could do complex life forms quickly. Using the classic pattern above and a wrapped world of 594,441 cells the best I could ever do was around 1,000 generations / sec.
Another simple pattern
**********
.
................*
.................**
................**.......**********
And gliders
........................*...........
......................*.*...........
............**......**............**
...........*...*....**............**
**........*.....*...**..............
**........*...*.**....*.*...........
..........*.....*.......*...........
...........*...*....................
............**......................