Twinning project: 13 Water client. Project completion time: 1 month (1: 3,1 prototype design, i.e. the herein; 3 implement the code, i.e. the next blog post). Pair teammates: zxr; I: zsy. Division: xr responsible for the design and prototype of landscaping, I was responsible for writing the blog and prototype prick.
To be continued.How long ah write more
Prototype tool
AxureRp. Easy to read, generous and elegant, easy to preview, but also release. Other prototype tool? I do not know, but AxureRp really have not used entirely for me this prototype design software people play.I think of painting a picture of the students, unfinished pictures map 233
The Prototype Analysis
section of writing
I will explain our design in the design specification.
Picture Effects
Pair discussion
PSP
| PSP2.1 | Personal Software Process Stages | Estimated time consuming (minutes) | The actual time-consuming (minutes) |
|---|---|---|---|
| Planning | plan | and | and |
| Estimate | Estimate how much time this task requires | and | and |
| Developm | Develop | and | and |
| Analysis | Needs analysis (including learning new technologies) | and | and |
| Design Spec | Generate design documents | and | and |
| Design Review | Design Review | and | and |
| Coding Standard | Code specifications (development of appropriate norms for the current development) | and | and |
| Design | Specific design | and | and |
| Coding | Specific coding | and | and |
| Code Review | Code Review | and | and |
| Test | Test (self-test, modify the code, submit modifications) | and | and |
| Reporting | report | and | and |
| Test Report | testing report | and | and |
| Size Measurement | Computing workload | and | and |
| Postmortem & Process Improvement Plan | Later summarized, and process improvement plan | and | and |
| total | and | and |
Design Notes
- basic needs
- login interface
- Displays the current state of war
- To display the result of the battle
- To show the details of the battle
- Displaying the Leaderboard
- demand analysis
- login interface
- Client name, maximum (small) and close buttons.
- Account number, password input text box.
- Login, Register button.
- Displays the current state of war
- Display and selection of hand
- The order of size (value) in the lower portion of the display page.
- Display a picture, the current picture is a network diagram. Single hand for the mouse over effects will rise, a fall of leaving.
- Select Click to select (tap and hold to drag can achieve continuous selection).
- Sandun display, select and modify
- Sandun order in the front, the left page is displayed.
- Sandun choice after selecting comply with the rules of the hand, click on the corresponding button, the selected card hand that is leaving the area to jump pier area.
- Providing a modified operation.
- Conversations feature
- In the lower right part of reality in the form of a list.
- Support choose the default text is sent, or send a self-written text, plans to add a certain expression.
- Display and selection of hand
- To display the result of the battle
- Style for the table
- Header including but not limited to: war time, the outcome of the opponent (1-3), the MVP authorities (i.e., winning player).
- Display Format
- Time: 24-hour clock (date). eg 13:14 (19.5.20)
- Outcome: win / loss (that is, the first non-negative).
- Opponent: Player nicknames.
- MVP: first innings of the corresponding player nickname.
- To show the details of the battle
- Query function (display the results as above)
- Time nodes: all, nearly a month, nearly a week, the same day.
- The outcome of the case: show win / loss bureau.
- Opponents: with a rival participative sessions.
- Query function (display the results as above)
- Leaderboard
- Integral with the corresponding ranking
- login interface
Challenges and Resolutions
- Description difficult
- AxureRp use
- Interactive interface design desired effect achieved with landscaping
- Try to resolve (corresponding to the sequence described in 1)
- Look own description blind spot function keys, the function keys can seem arbitrary point will again have a point
At least I was so dry。 - Realization: AxureRp interactive button. Beautification: network map
Maybe on their own P。
- Look own description blind spot function keys, the function keys can seem arbitrary point will again have a point
- Is resolved
- They are good to solve the difficulties.
- What harvest
- Have some of the features of AxureRp.
- Comfortable after some interactive realize discussed.
Experience (zsy):
This weekend (19.9.21-19.9.22) really did not do anything prototype design, Friday (19.9.20) teammates decided to use AxureRp, he himself also downloaded messing around a bit. Saturday (19.9.21) morning, I finished the login screen and stage victories to the interface, to pay the remainder of his teammates realized. After I went to prepare other matters, after the end of Sunday (19.9.22), and found it changed my teammates login screen, change was really nice, I feel able to release, if Gaga animation or background, just like a page tour login screen. Past victories and are still used in my design. As for the rest of his teammates spent most of the time the current state of play against the design, he adds background and some action interaction, I do not know whether he was satisfied, at least I was satisfied, feeling almost beautify related Landlords .
AxureRp study, mainly the mouse over the icon to see tips, harmless to humans and animals feel on the point of the play, I just do not what wonderful things are operating, Gaga interactive and connected feel done, but fortunately I Battle interface did not do, I guess even the brand is not centered tone will goStopping when he discovered his teammates in tune 233.
The prototype design I basically did what force, but I have a lot of mouth Hu desired effect, also had a discussion with his teammates, he felt feasible also draw outI have the potential to Party!. As for after the code implementation, UI teammates and I will certainly write together, as for algorithm design, preliminary currently divided into two parts: written teammate Paul at the end algorithms (should choose greedy), in order to ensure that the final to achieve complete processing input output. I went to Mang wave of other algorithms to expect to achieve better results.
This part of the algorithm talk about my ideas.
If misunderstood, please correct me Tell me.
In fact, there is a simple method to achieve, to change the so-called machine learning to match the depth of learning just fine. As far as technical assistant, said after the program we've written, there will be 72 hours of training game, which means I can go blind play, then remember that each of the four results of the hand combination, the last four kinds of distribution players way corresponding score normalization, when the last test, I changed Learning is Matching, Chou Chou my experience that is set inside the hand combination of whether there have been, if the highest score over how to combine the emergence of direct generation of correspondence distribution. The problem is that a lot of water 13 Collection, entirely possible I never get over the combination (in fact, if we give up considering color, I think 72 hours Game + data generated on their own intuition should be able to get most of the hand combination )NaiveBayesian ???. Why learn ah, it is because we can not get the complete works ah, or fees that time Zuoshen, matching fly more than learningIf you are out of the examination of the original title, not seconds to write?. However, this method is the difficulty of the possible consequences of the expected results, because you might get your hands played, but there are other combinations of other people, a bad luck situation, your decision may have been the highest score loseSo to learn ah.
A topic, K boss mentioned reinforcement learning, and indeed for this information game, reinforcement learning is a weapon. On my foundation reinforcement learning only remain in Wsp teacher of a Q-Learning courses, and then discuss xg Q-Learning as if that is a spfa, he asked me why can not run dij, then I do not know, I said I I do not know, now I probably know. Because dij is there from loose push heap (heap optimized version of dij), and Q-Learning can also previous knowledge (I think that is not relaxed distance) to learn, which is obviously not dij, but Why this makes Q-Learning can work, and I just stay on perceptions. After last Tuesday (19.9.17) to see the water is 13, I think the second picture below (my source: Almost Human).
& emspl apparently inherent information 13 of water, is a rectangle. But I think it and Go poker and a very important difference is: one-time decision-making stage. That is, after you have finished selecting Sandun, can not be changed after the cards, direct size comparison with the rest of the players. Go and poker and there is a current state to dictate your decision now, while 13 water is more like a decision, which is the reason greedy choice, it seems difficult for us to go to decision-making in the hands of others possible (because one-off decisions, and may not even guess of course, that there is no chess process).
Secondly, I think the idea of TF-IDF algorithm. If my current hand (or some combination thereof) of the TF-IDF index high, indicating complete works in hand, the pair of cards (or some combination thereof) has a good distinguish it from other card type ability, then the deck is likely to have some kind of character worthy of our attention.Dim ah, ah can not think
In summary, in the mouth HuIf I say will be able to achieve, I was immortal words?I did not write one line of code algorithm, look I can write it after learning. Did not intend to learn so early reinforcement learning, since met, see if you can do it, although I think if a comprehensive study, then three weeks really too short (theoretical basisText description seems to have lost its soul, Non-real transfer packagegym stick stand how I feel super silly，理解后进一步运用）。所以打算选择一种折中的方式实现这次作业，如果我真的实现了一个效果不错的算法，相信对我是一个莫大的鼓励。怎么定义不错？至少面对sort能够有较高的胜率吧