Jim Fleming (@jimmfleming) header
Jim Fleming (@jimmfleming) profile picture

Doing things with ML

San Francisco, CA
Joined December 2007
Tracing the front wheel of the bicycle over several episodes. I suspect the steering column is a little stiff since the dynamics look somewhat different from the paper.
1
7
Essential reading (2016) to get started with self-play ML using TensorFlow. The leap from backgammon to chess/go was astonishing. From there to sport simulation is closer than you think teapawt.net/jimmfleming/status/717…
1
1
8
That reads like a queued tweet from December. Especially in contrast to FB blocking crypto ads yesterday, which feels like the better move: teapawt.net/jack/status/9587432385…
Instant buying (and selling, if you don’t want to hodl) of Bitcoin is now available to most Cash App customers. We support Bitcoin because we see it as a long-term path towards greater financial access for all. This is a small step. We also made this! cash.me/bitcoin
4
I see a looming AI creativity copyright crisis ahead! (@ a talk at Harvard on the famous 'monkey selfie' case).
1
1
9
I'll need to perturb the bike at the beginning of the episodes, otherwise it's too easy:
3
7
I got the bike hooked up to Bullet and parts of the Gym env setup too!
1
5
22
I wanted a fun weekend project so I'm recreating the env from: "It Takes Two Neurons to Ride a Bicycle": citeseerx.ist.psu.edu/viewdo…
3
2
3
20
Reading group for AlphaGo Zero paper meetup.com/deep-learning-sf/…
2
9
blog post by Sanjay Krishnan: Learning Long Duration Sequential Task Structure From Demonstrations bair.berkeley.edu/blog/2017/…
11
39
Useful guide to the Gumbel-max trick from @xtimv buff.ly/2hgy1dX #MachineLearning
1
5
20
Simple guide to interpreting random forests from Pivotal buff.ly/2xl1wob #MachineLearning
1
5
New post from BAIR on their paper "Learning to Optimize" buff.ly/2jEYHcY #DeepLearning #MachineLearning
10
29
Jim Fleming retweeted
Machine learning is a big deal -- John Deere is paying $305 million to buy a 60-person Silicon Valley tech startup axios.com/john-deere-is-payi…
1
17
1
35
When you get rounding errors in your package versions: "html5lib 0.999999999" pypi.python.org/pypi/html5li… – at San Francisco, CA
6
Goog extends 'Wide and Deep' to higher order feature interactions. Notes from @jimmfleming's #deeplearningsf my.memo.ai/external/vJ-iBYK4…
8
1
27
"The Business of Artificial Intelligence" from @erikbryn buff.ly/2f2qh20
1
6
"The Bias-Variance Dilemma" from Daniel Geng and Shannon Shih: buff.ly/2uP1t2N #machinelearning
9
"Bayesian Neural Networks with Random Inputs for Model Based RL" from @TDataScience buff.ly/2vmXyvB #reinforcementlearning
4
18