pow(a,b) not supported as such

[rll2021]

The code assumes that binary operators are infix. No prefix binary operator is handled correctly, viz., 'atan2(a,b)', 'pow(a,b)'. While atan2 is noted as "NOT supported at this time" in file operators_list.txt, we do not so denote pow.

• I propose to add 'pow(a,b)' to the "NOT supported" list, with a parenthetical note that the infix form (a**b) is supported.

Additionally, 'pow' appears twice in the set 'operators' in base_class.py, first as "ast.Pow: tf.pow, # e.g., a ** 2", and later as "'pow': tf.pow, # e.g., pow(a, b)".

• I propose striking the latter.

[asksak]

The code assumes that binary operators are infix. No prefix binary operator is handled correctly, viz., 'atan2(a,b)', 'pow(a,b)'. While atan2 is noted as "NOT supported at this time" in file operators_list.txt, we do not so denote pow.

• I propose to add 'pow(a,b)' to the "NOT supported" list, with a parenthetical note that the infix form (a**b) is supported.

Additionally, 'pow' appears twice in the set 'operators' in base_class.py, first as "ast.Pow: tf.pow, # e.g., a ** 2", and later as "'pow': tf.pow, # e.g., pow(a, b)".

• I propose striking the latter.

Hello,

a^b is a good idea, I've been looking for operators to add.

Will do it and post the method.

Regards

Aymen

[kstaats]

Richard, Do you have a revised version for me to test? Cheers, kai

…

On 4/19/21 11:24 AM, Richard wrote: The code assumes that binary operators are infix. No prefix binary operator is handled correctly, viz., 'atan2(a,b)', 'pow(a,b)'. While atan2 is noted as "NOT supported at this time" in file operators_list.txt, we do not so denote pow. - I propose to add 'pow(a,b)' to the "NOT supported" list, with a parenthetical note that the infix form (a**b) is supported. Additionally, 'pow' appears twice in the set 'operators' in base_class.py, first as "ast.Pow: tf.pow, # e.g., a ** 2", and later as "'pow': tf.pow, # e.g., pow(a, b)". - I propose striking the latter.

[rll2021]

No, I do not, but the suggestion I made above is a two line fix.
I have not heard from Asksak, so I don't know if he has anything.

[asksak]

No, I do not, but the suggestion I made above is a two line fix.
I have not heard from Asksak, so I don't know if he has anything.

Hello,

The first (pow) is ** in operators and is x^2, while the second (pow) is pow in operators which is a^b so I did this in labels:

def fx_eval_label(self, tree, node_id):
	
		'''
		Evaluate all or part of a Tree (starting at node_id) and return a raw mutivariate expression ('algo_raw').
		
		This method is called once per Tree, but may be called at any time to prepare an expression for any full or 
		partial (branch) Tree contained in 'population'. Pass the starting node for recursion via the local variable 
		'node_id' where the local variable 'tree' is a copy of the Tree you desire to evaluate.
		
		Called by: fx_eval_poly, fx_eval_label (recursively)
		
		Arguments required: tree, node_id
		'''
		
		# if tree[6, node_id] == 'not': tree[6, node_id] = ', not' # temp until this can be fixed at data_load
		
		node_id = int(node_id)
		
		if tree[8, node_id] == '0': # arity of 0 for the pattern '[term]'
			return tree[6, node_id] # 'node_label' (function or terminal)
			
		else:
			if tree[8, node_id] == '1': # arity of 1 for the explicit pattern 'not [term]'
				return tree[6, node_id] + self.fx_eval_label(tree, tree[9, node_id])	 # rll 20210201

			elif tree[8, node_id] == '2': # arity of 2 for the pattern '[func] [term] [func]'
				if tree[6, node_id] == 'min':
					return tree[6, node_id]  + self.fx_eval_label(tree, tree[9, node_id]) + self.fx_eval_label(tree, tree[10, node_id]) #asksak
				if tree[6, node_id] == 'max':
					return tree[6, node_id]  + self.fx_eval_label(tree, tree[9, node_id]) + self.fx_eval_label(tree, tree[10, node_id]) #asksak
				if tree[6, node_id] == 'pow':
					return tree[6, node_id]  + self.fx_eval_label(tree, tree[9, node_id]) + self.fx_eval_label(tree, tree[10, node_id]) #asksak, could combine all three lines with OR
				else:
					return self.fx_eval_label(tree, tree[9, node_id]) + tree[6, node_id] + self.fx_eval_label(tree, tree[10, node_id])	#asksak
				
			elif tree[8, node_id] == '3': # arity of 3 for the explicit pattern 'if [term] then [term] else [term]'
				# This fails in sympify. rll 20210206
				#This needs a new set of code. #asksak
				return tree[6, node_id] + self.fx_eval_label(tree, tree[9, node_id]) + ' then ' + self.fx_eval_label(tree, tree[10, node_id]) + ' else ' + self.fx_eval_label(tree, tree[11, node_id])

Note THE ARITY:

operator, arity
+,2
-,2
*,2
/,2
**,2
min,2
max,2
pow,2
sqrt,1
abs,1
log,1
cos,1
sin,1
exp,1
expm1,1

I will add all the code when I learn where and how to upload on GitHub.

As a side note (so far)

1. I added min, max
2. I fixed arity
3. I modified fx_eval_label for compatibility
4. I added dynamic mutation and standard mutation, both selectable at program startup
5. I optimized memory utilization of the program (collected garbage)
6. I optimized variable assignment
7. I optimized loops
8. I removed all '(' and ',' from fx_eval_label which I had added but turned out to interfere with sympy, kstaats version was better.
9. I tested all the above with known test functions and all results solved in a few generations.

Will update u soon

[kstaats]

Thank you for your continued work on Karoo. My time has been very limited but I am eager to get back to this project and incorporate your good work into the next version. I will reach out to you, off of this thread, and ask for a summary of all the changes you have found most valuable.

…

On 6/17/21 1:35 AM, asksak wrote: > No, I do not, but the suggestion I made above is a two line fix. > I have not heard from Asksak, so I don't know if he has anything. Hello, The first (pow) is ** in operators and is x^2, while the second (pow) is pow in operators which is a^b so I did this in labels: ``` def fx_eval_label(self, tree, node_id): ''' Evaluate all or part of a Tree (starting at node_id) and return a raw mutivariate expression ('algo_raw'). This method is called once per Tree, but may be called at any time to prepare an expression for any full or partial (branch) Tree contained in 'population'. Pass the starting node for recursion via the local variable 'node_id' where the local variable 'tree' is a copy of the Tree you desire to evaluate. Called by: fx_eval_poly, fx_eval_label (recursively) Arguments required: tree, node_id ''' # if tree[6, node_id] == 'not': tree[6, node_id] = ', not' # temp until this can be fixed at data_load node_id = int(node_id) if tree[8, node_id] == '0': # arity of 0 for the pattern '[term]' return tree[6, node_id] # 'node_label' (function or terminal) else: if tree[8, node_id] == '1': # arity of 1 for the explicit pattern 'not [term]' return tree[6, node_id] + self.fx_eval_label(tree, tree[9, node_id]) # rll 20210201 elif tree[8, node_id] == '2': # arity of 2 for the pattern '[func] [term] [func]' if tree[6, node_id] == 'min': return tree[6, node_id] + self.fx_eval_label(tree, tree[9, node_id]) + self.fx_eval_label(tree, tree[10, node_id]) #asksak if tree[6, node_id] == 'max': return tree[6, node_id] + self.fx_eval_label(tree, tree[9, node_id]) + self.fx_eval_label(tree, tree[10, node_id]) #asksak if tree[6, node_id] == 'pow': return tree[6, node_id] + self.fx_eval_label(tree, tree[9, node_id]) + self.fx_eval_label(tree, tree[10, node_id]) #asksak, could combine all three lines with OR else: return self.fx_eval_label(tree, tree[9, node_id]) + tree[6, node_id] + self.fx_eval_label(tree, tree[10, node_id]) #asksak elif tree[8, node_id] == '3': # arity of 3 for the explicit pattern 'if [term] then [term] else [term]' # This fails in sympify. rll 20210206 #This needs a new set of code. #asksak return tree[6, node_id] + self.fx_eval_label(tree, tree[9, node_id]) + ' then ' + self.fx_eval_label(tree, tree[10, node_id]) + ' else ' + self.fx_eval_label(tree, tree[11, node_id]) ``` Note THE ARITY: operator, arity +,2 -,2 *,2 /,2 **,2 min,2 max,2 pow,2 sqrt,1 abs,1 log,1 cos,1 sin,1 exp,1 expm1,1 I will add all the code when I learn where and how to upload on GitHub. As a side note (so far) 1. I added min, max 2. I fixed arity 3. I modified fx_eval_label for compatibility 4. I added dynamic mutation and standard mutation, both selectable at program startup 5. I optimized memory utilization of the program (collected garbage) 6. I optimized variable assignment 7. I optimized loops 8. I removed all '(' and ',' from fx_eval_label which I had added but turned out to interfere with sympy, kstaats version was better. 9. I tested all the above with known test functions and all results solved in a few generations. Will update u soon